API - Technical Report 941 Sept. 2008 - The Technical Basis Document for API RP 941

The Technical Basis Document for API RP 941

API TECHNICAL REPORT 941SEPTEMBER 2008

The Technical Basis Document for API RP 941

Downstream Segment

API TECHNICAL REPORT 941SEPTEMBER 2008

Special Notes

API publications necessarily address problems of a general nature. With respect to particular circumstances, local,state, and federal laws and regulations should be reviewed.

Neither API nor any of API's employees, subcontractors, consultants, committees, or other assignees make anywarranty or representation, either express or implied, with respect to the accuracy, completeness, or usefulness of theinformation contained herein, or assume any liability or responsibility for any use, or the results of such use, of anyinformation or process disclosed in this publication. Neither API nor any of API's employees, subcontractors,consultants, or other assignees represent that use of this publication would not infringe upon privately owned rights.

Users of this technical report should not rely exclusively on the information contained in this document. Soundbusiness, scientific, engineering, and safety judgment should be used in employing the information contained herein.

API publications may be used by anyone desiring to do so. Every effort has been made by the Institute to assure theaccuracy and reliability of the data contained in them; however, the Institute makes no representation, warranty, orguarantee in connection with this publication and hereby expressly disclaims any liability or responsibility for loss ordamage resulting from its use or for the violation of any authorities having jurisdiction with which this publication mayconflict.

API publications are published to facilitate the broad availability of proven, sound engineering and operatingpractices. These publications are not intended to obviate the need for applying sound engineering judgmentregarding when and where these publications should be utilized. The formulation and publication of API publicationsis not intended in any way to inhibit anyone from using any other practices.

Any manufacturer marking equipment or materials in conformance with the marking requirements of an API standardis solely responsible for complying with all the applicable requirements of that standard. API does not represent,warrant, or guarantee that such products do in fact conform to the applicable API standard.

All rights reserved. No part of this work may be reproduced, stored in a retrieval system, or transmitted by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior written permission from the publisher. Contact the Publisher, API

Publishing Services, 1220 L Street, N.W., Washington, D.C. 20005.

Copyright © 2008 American Petroleum Institute

Foreword

Nothing contained in any API publication is to be construed as granting any right, by implication or otherwise, for themanufacture, sale, or use of any method, apparatus, or product covered by letters patent. Neither should anythingcontained in the publication be construed as insuring anyone against liability for infringement of letters patent.

This document was produced under API standardization procedures that ensure appropriate notification andparticipation in the developmental process and is designated as an API standard. Questions concerning theinterpretation of the content of this publication or comments and questions concerning the procedures under whichthis publication was developed should be directed in writing to the Director of Standards, American PetroleumInstitute, 1220 L Street, N.W., Washington, D.C. 20005. Requests for permission to reproduce or translate all or anypart of the material published herein should also be addressed to the director.

Generally, API standards are reviewed and revised, reaffirmed, or withdrawn at least every five years. A one-timeextension of up to two years may be added to this review cycle. Status of the publication can be ascertained from theAPI Standards Department, telephone (202) 682-8000. A catalog of API publications and materials is publishedannually and updated quarterly by API, 1220 L Street, N.W., Washington, D.C. 20005.

Suggested revisions are invited and should be submitted to the Standards Department, API, 1220 L Street, NW,Washington, D.C. 20005, [email protected].

iii

Report to API - The Technical Basis For API RP 941

TABLE OF CONTENTS

0HEXECUTIVE SUMMARY FOR THE TECHNICAL BASIS DOCUMENT ..................................................... 361H1 1H1.0 ABSTRACT ..................................................................................................................................... 362H5

2H2.0 INTRODUCTION .............................................................................................................................. 363H6 3H2.1 What is Hydrogen Attack? ............................................................................................................ 364H7 4H2.2 What is Fugacity and Why Is It Important .................................................................................. 365H11 5H2.3 Details of the Attack Progression .............................................................................................. 366H12

6H3.0 ATTACK OF MATERIALS CONTAINING FE3C ........................................................................... 367H20

7H4.0 METHANE PRESSURE RELATIONS WITH CEMENTITE .......................................................... 368H23

8H5.0 STEELS WITH MORE STABLE CARBIDES ................................................................................ 369H25

9H6.0 BEHAVIOR OF STEELS WITH COMPLEX ALLOY CARBIDES ................................................. 370H26

10H7.0 RESEARCH STUDIES .................................................................................................................. 371H27

11H8.0 STUDIES ........................................................................................................................................ 372H28

12H9.0 EFFECT OF STRESS .................................................................................................................... 373H32

13H10.0 EFFECT OF TEMPERATURE ....................................................................................................... 374H33

14H11.0 EFFECT OF HYDROGEN PRESSURE ........................................................................................ 375H34 15H11.1 Materials Issues ........................................................................................................................... 376H34

16H12.0 EXPOSURE TIME OR STRESS .................................................................................................... 377H35

17H13.0 EUROPEAN STUDIES .................................................................................................................. 378H36

18H14.0 NEW ANALYSIS OF MPC WORK ................................................................................................ 379H40

19H15.0 DEVELOPMENT OF P/T LIMITS NEAR THE CREEP RANGE AND ABOVE ............................ 380H42

20H16.0 A TECHNICAL BASIS ................................................................................................................... 381H43

21H17.0 DISCUSSION ................................................................................................................................. 382H45

22H18.0 FINAL COMMENT ......................................................................................................................... 383H47

23H19.0 OVERVIEW AND CONCLUSION .................................................................................................. 384H49

24H20.0 REFERENCES ............................................................................................................................... 385H51

25H21.0 FIGURES ....................................................................................................................................... 386H56 26HAPPENDIX A – BACKGROUND TO THE NELSON CURVES ............................................................... 387H124

27H1.0 HISTORY ..................................................................................................................................... 388H125

28H2.0 WHAT DATA IS AVAILABLE? ................................................................................................... 389H128

29H3.0 WHERE DID THE DATA COME FROM? .................................................................................... 390H129 30H3.1 API Publication 941 Fifth Edition Supplement 1 April 1998 ................................................ 391H129 31H3.2 Figure 1 Points C-0.5Mo Appendix A ..................................................................................... 392H130 32H3.3 File Correspondence but Points Not Plotted .......................................................................... 393H135 33H3.4 Points Plotted for 1-1/4Cr-1/2Mo and 2-1/4Cr-1Mo ................................................................. 394H138 34H3.5 Attack / No Attack Points .......................................................................................................... 395H139

35H4.0 DISCUSSION ............................................................................................................................... 396H141

36H5.0 CONCLUSIONS ........................................................................................................................... 397H145

37H6.0 FIGURES ..................................................................................................................................... 398H146


38HAPPENDIX B – COMMONLY ASKED QUESTIONS ............................................................................. 399H157

39H1.0 HOW DO WE GET CARBON ACTIVITY AND CONTENT? ....................................................... 400H158 40H1.1 How Much Carbon Is Necessary To Exhaust The Carbide Formers? .................................. 401H161

41H2.0 ARE THE API RP 941 CURVES WHERE THEY BELONG? ..................................................... 402H162

42H3.0 WHAT VARIABLES MIGHT ADVERSELY INFLUENCE BEHAVIOR IN LONG TERM SERVICE? ................................................................................................................................... 403H163

43H4.0 WHAT OTHER FACTORS MIGHT PLAY A ROLE IN THE APPEARANCE OF HYDROGEN ATTACK? .................................................................................................................................... 404H164

44H5.0 WHY IS HYDROGEN ATTACK SEEMINGLY UNPREDICTABLE? .......................................... 405H165

45H6.0 WHY HAVEN’T THE NECESSARY CRITICAL EXPERIMENTS BEEN RUN? ......................... 406H166

46H7.0 WHAT IS ACTUALLY HAPPENING DURING EARLY STATES OF HYDROGEN ATTACK? . 407H167

47H8.0 REFERENCES ............................................................................................................................. 408H168 APPENDIX C – ESTIMATING DAMAGE RATES FOR LIFE ASSESSMENT ........................................ 169 48HAPPENDIX D – EFFECTIVE PRESSURES OF HYDROGEN IN STEEL COVERED BY

CLAD/OVERLAY AND/OR CORROSION PRODUCT .................................................. 409H178 49HAPPENDIX E – OBSTACLES TO UNDERSTANDING HTHA ................................................................ 410H184 50HAPPENDIX F – APPLICATION AND SUMMARY OF PW PARAMETRIC MODEL ................................ 411H187

51H1.0 SUMMARY OF THEORETICAL KINETIC TREATMENTS ON BUBBLE GROWTH ................. 412H189 52H1.1 Diffusion Models - Grain Boundary and Surface Diffusion ................................................... 413H189 53H1.2 Coupled Diffusion and Creep Model ........................................................................................ 414H189

54H2.0 RELATIONSHIP BETWEEN THEORETICAL EQUATIONS AND ARRHENIUS TYPE EXPRESSIONS ........................................................................................................................... 415H191

55H2.1 Grain Boundary Diffusion (GBD) Process ............................................................................... 416H191 56H2.2 Surface Diffusion (SD) Process ................................................................................................ 417H192 57H2.3 Coupling Between GBD and SD Processes ............................................................................ 418H192 58H2.4 Power-Law Creep Process ........................................................................................................ 419H192 59H2.5 General Form of All Processes ................................................................................................. 420H192

60H3.0 DERIVATION OF ENGINEERING PARAMETER PW ................................................................ 421H194

61H4.0 EXAMPLE OF EXPERIMENTAL RESULTS RELATIVE TO THE ARRHENIUS TYPE EXPRESSION .............................................................................................................................. 422H196

62H5.0 APPLICATION OF THE PW PARAMETER – PREDICTION OF TIME DEPENDENT CRITICAL CURVES ...................................................................................................................................... 423H198

63H5.1 Discussion of the Values of and Q ....................................................................................... 424H198 64H5.2 Discussion of API RP 941 versus Parametric Equations for Several Materials .................. 425H200

65H6.0 APPLICATION OF THE PW PARAMETER – ESTIMATION OF THE EFFECT OF APPLIED STRESS ....................................................................................................................................... 426H205

66H6.1 Derivation of Pw Including the Stress ..................................................................................... 427H205 67H6.2 Effect of Applied Stress on the Critical Curve of 2.25Cr-1Mo Steel ..................................... 428H206

68H7.0 APPLICATION OF THE PW PARAMETER – ESTIMATION OF EFFECT OF OVERLAY ........ 429H207 69H7.1 Hydrogen Distribution in a Vessel Wall with Overlay During Operation .............................. 430H207 70H7.2 The Effect of Overlay on the Critical Curve ............................................................................. 431H208

71H8.0 CONCLUSIONS ........................................................................................................................... 432H210

72H9.0 REFERENCES ............................................................................................................................. 433H211

73H10.0 FIGURES ..................................................................................................................................... 434H212


74HAPPENDIX G – OVERVIEW OF EUROPEAN RESEARCH .................................................................. 435H231

75H1.0 INTRODUCTION .......................................................................................................................... 436H232

76H2.0 NON-UNIFORM HYDROGEN ATTACK CAVITATION AND THE ROLE OF INTERACTION WITH CREEP ............................................................................................................................... 437H235

77H2.1 Introduction ................................................................................................................................ 438H235 78H2.2 Polycrystal Model for HA .......................................................................................................... 439H236 79H2.3 Method of Analysis .................................................................................................................... 440H242 80H2.4 Discussion .................................................................................................................................. 441H248 81H2.5 Conclusion .................................................................................................................................. 442H250

82H3.0 COMPARISON OF DELFT AND APPROACH REPORTED FROM JAPAN ............................. 443H252 83H3.1 Introduction to the Hydrogen Attack Models .......................................................................... 444H252 84H3.2 Calculation of the Methane Pressure ....................................................................................... 445H253 85H3.3 Comparison Between Methane Pressures .............................................................................. 446H254 86H3.4 Description of Void Growth Due to Diffusion.......................................................................... 447H255 87H3.5 Comparison of Predicted Void Growth Due To Diffusion ...................................................... 448H256 88H3.6 Creep As An Additional Deformation Mechanism .................................................................. 449H258 89H3.7 Conclusions ................................................................................................................................ 450H260

90H4.0 FIGURES ..................................................................................................................................... 451H261

91H5.0 REFERENCES ............................................................................................................................. 452H292 92HAPPENDIX H – SAMPLE PROBLEMS UTILIZING APPENDIX C AND DISCUSSION DEVELOPED BY

SOME COMMITTEE MEMBERS DURING REVIEW OF THE REPORT ...................... 453H294

93H1.0 ESTIMATING DAMAGE RATES FOR LIFE ASSESSMENT: BASIC METHODOLOGY .......... 454H295

94H2.0 EXAMPLES: HYPOTHETICAL CASES USING APPENDIX C .................................................. 455H296 95H2.1 Case 1: Carbon Steel channel flange and shell welds discovered in 1.25Cr exchanger ... 456H296 96H2.2 Case 1A: 225 psia at 620oF ....................................................................................................... 457H296 97H2.3 Case 1B: 225 psia at 750oF ........................................................................................................ 458H297 98H2.4 Case 2: Carbon steel pipe ......................................................................................................... 459H297 99H2.5 Case 3: C-0.5Mo exchanger channel cover ............................................................................. 460H297 100H2.6 Case 4: C-0.5Mo Methanator Short-Time Excursion ............................................................. 461H298 101H2.7 Case 5: Mn-0.5Mo exchanger with austenitic weld overlay/clad .......................................... 462H299 102H2.8 Case 5A: Consider Continued Operation at Current Conditions .......................................... 463H299 103H2.9 Case 5B: Operations wants to increase the operating conditions ....................................... 464H300 104H2.10 Case 6: 1Cr catalytic reformer reactor .................................................................................... 465H300 105H2.11 Case 7: Over temperature Conditions in a 2.25Cr 1 Mo Reactor ......................................... 466H300


1

EXECUTIVE SUMMARY FOR THE TECHNICAL BASIS DOCUMENT

Before the first edition of API Publication 941 “Steels for Hydrogen Service at Elevated Temperatures and Pressures in Petroleum Refineries and Petrochemical Plants” appeared in 1970, there had been fundamental questions regarding the technical basis for the materials performance curves contained in the document (1-6). Based upon sparse laboratory data combined with plant experience, with only a few exceptions, the curves have done an exceptionally good job at safely directing the refining industry in selecting materials based upon operating temperature, hydrogen partial pressure, and the metallurgy of the equipment being considered. However, in some cases, past editions of the API RP 941 document were of limited value, most notably for C-0.5Mo material. Today, with refining plants aging, engineers are seeking assurances that the curves are suitable for predicting continuing satisfactory performance for decades into the future. Of concern also, is the unusual shape of the 1.25Cr curve which appears inconsistent with the other curves without any obvious technical reason. Most important, engineers require technical justification for decisions made regarding suitability for service after process excursions that exceed the 941 “safe” limits. The API Subcommittee for Corrosion and Materials commissioned the 941 Task Group to provide a technical basis document that goes beyond empirical evidence to address three issues: 1. Do the curves given in the current edition have the correct shapes and locations? 2. Are the curves likely to “change” with time as our plants become older? 3. What methodology and data can be used to handle process excursions? The Task Group met these objectives. The Technical Basis Document (TBD) that follows is the result of several years of effort incorporating the technical insights contributed by participating Japanese and European specialists with those presented to API from the United States. Dr. Martin Prager of The Materials Properties Council Inc. (MPC) prepared much of the TBD and developed the approach set forth in the main body of this document. Please see the acknowledgements of the contributions of the others noted below. Details can be found in the respective Appendix sections It is important in considering this work that it is a research report, not a recommended practice. Those workers most closely involved in this report believe some of the findings are so well supported that they can be immediately brought into the next edition of the 941 Recommended Practice. Examples are identified below. Other findings push the edge of our understanding and give very useful insight without yet being RP-ready. This is hard work, the complexity of which is matched by a frustrating lack of quality data in many cases. What goes into the next API RP 941 will be the work of the overall 941 Task Group. Immediately below are highlights of the attached Technical Basis Document. For many, this level of detail along with selected portions of the document to handle specific issues will suffice. Those who delve into the entire document will be rewarded by Dr. Prager‟s elegant explanations that bring us down to the mechanistic level, and back out again to provide guidance in dealing with actual plant challenges.


2

Highlights of the Draft Technical Basis Report for API RP 941 1.) The shapes and locations of the curves in the most recent RP 941 are essentially correct.

a) The carbon steel curve appears to be perhaps 30oF to as much as 50oF conservative. However, there is insufficient laboratory or plant data to justify adjusting the curve, and it is important to note that under special circumstances (e.g., unusually high stresses) there have been failures even below the current carbon steel curve.

b) All curves should have essentially the same shape. Therefore the unusual “kink” in the 1.25Cr curve is likely incorrect. The 941 Task Group should consider adjusting this curve.

c) The shape of the curves, where they go essentially “vertical” at low hydrogen partial pressures and become almost flat at high pressures, can be understood by taking into account both kinetics, thermodynamics and materials‟ strength. To a remarkable extent, attack quantitatively tracks the hydrogen and carbon solubility, which are low at low temperatures. Nevertheless, at low temperatures the methane pressure formed from even small concentrations of these elements can be enormous, much higher than the strength of the materials. Fortunately the kinetics are very slow at low temperatures, so that this full pressure is unlikely to be realized.

d) A key difference among alloys is actually carbon activity, which largely comes down to the amount of free carbon left in solid solution, and that is in equilibrium with the carbides. The predominate carbide in carbon steel is Cementite (Fe3C). Cementite is the least stable carbide and, when it is present, provides much of the easily reacted carbon. The relative stability of the carbides found in various alloys is discussed in this TBD.

2.) We found no data or theory that would cause significant concern with equipment in hydrogen service operating below the current edition curve limits even for many hundreds of thousands of hours, as long as the equipment operates under code stress limits.

a) However, there is at least a theoretical concern that if the equipment is operated above the curve for that metallurgy, then hydrogen attack may initiate and possibly even continue after the operating conditions are returned to below the curve.

b) There is a theoretical basis for believing that for all practical purposes below certain conditions hydrogen attack will never occur.

3.) The work found that the “incubation curves” as given in the current and previous editions are likely to be correct only for the specific set of conditions used to develop the curves. Yet we know that significant attack does not occur the instant material is exposed to conditions above the curves. An alternative approach for handling process excursions which is better founded on reaction rate and material strength principles is given in this report. Worked examples are given in Appendix C of the report. Essentially:

a) High temperature hydrogen attack will not occur as long as the creep strength of the material is greater than the internal pressures caused by the buildup of methane.

b) The amount of methane pressure buildup in the steel depends upon the hydrogen pressure of the process, the temperature, and carbon activity in the material.

c) The greater the hydrogen pressure of the process, the more methane will be formed, but because at higher pressures, the gases no longer perform ideally, the attack becomes less sensitive to pressure and the curves tend to flatten at higher pressures. More importantly, the attack follows quantitatively the carbon and hydrogen availability in the material, and at lower temperatures the carbon and hydrogen solubility significantly decrease. Therefore, at lower


3

temperatures, the hydrogen pressure must be greater to produce the destructive methane pressure in a given period of time.

d) Higher temperatures will actually decrease the potential maximum methane pressure, but at the same time, higher temperatures reduce the creep resistance of the material and increase reaction rates.

e) The material, and specifically the carbon and carbide content, is important. The greater free carbon that is available, the more methane pressure will build up for a set of operating conditions. A thermodynamically unstable carbide, such as Fe3C, will actually worsen the situation by allowing more carbon to become available. Stable carbides, as found in low alloy material, provide markedly improved resistance by both reducing the amount of carbon available, and at the same time increasing the creep strength.

f) Austenitic and even ferritic cladding dramatically reduces the effective hydrogen partial pressure behind the cladding, and can greatly retard or prevent HTHA. The attached report provides useful curves to easily predict the benefits of cladding. We believe this benefit is now well enough established that it should be recognized in the next edition of API RP 941.

Acknowledgements The Task Group is particularly appreciative of the analysis and contributions of Dr. Tadamichi Sakai and Dr. Tohru Nomura who actively participated in Task Group meetings reporting on cooperative work funded by the Petroleum Energy Center of Japan through a Task group of the Japan Petroleum Institute. They built on work contributed by Japan Pressure Vessel Research Council as presented by Sakai and Nomura in Appendix F. Also, the work of European investigators H. van Wortel of TNO, S. Schlogl and E. van der Geissen of TU Delft and T. Manolatis and A. Baker of JRC Petten as communicated to the Task Group by one member (R. Koers) was most enlightening and is contained in Appendix G. Finally, Jay Cantwell‟s wealth of personal experience and insights gained in the refining industry (Appendix A) provided an invaluable starting point. E. H. (Ned) Niccolls


4

REPORT TO API - THE TECHNICAL BASIS FOR API RP 941

5

1.0 ABSTRACT

Reports covering a half-century of comprehensive research on hydrogen attack have been reviewed. The major investigators were found to agree about what information would be needed to model the curves presented in API RP 941. However, they concluded that quantification of key, very complex material property and performance inputs is not possible. Prediction of attack limits from first principles therefore remains elusive. With the benefit of hindsight, the curves in API RP 941 are explained herein. A series of reasonable assumptions appear to justify Nelson's placement of the lines for carbon and low alloy steels. The approach proposed here is applied to these common steels and agrees with trends in attack thresholds established by experience. It is based on the obvious and long-held notions that if the methane pressure in voids is low compared to the material‟s strength or methane forming reaction rates are low, attack does not occur. The approach is flexible and can be applied to all carbon and low alloy steels. It can also be used as a starting point to estimate the effect of applied stress on time-dependent behavior. Application of these models to refinery equipment, especially clad components, has been attempted and the results are credible. Ferritic and austenitic stainless steel overlay and cladding are clearly effective. However, practical implementation of the principles is impeded by uncertainties regarding diffusivity, solubility, absorption rates, and fluxes of hydrogen and the effects of stress and materials strength.

Among the stumbling blocks to successful modeling of hydrogen attack is the lack of knowledge of relevant concentrations and activities of carbon and alloy elements remaining in solution after heat treatment. Also, there is scant knowledge of details about void nucleation and the rates of the methane forming reactions in voids. Local compositions at grain boundaries and the compositions of carbides are probably important, but are not known with certainty. The manner and rate of the evolution of hydrogen attack damage have not been studied quantitatively. Prediction of attack boundaries is difficult since materials of a grade differ in critical respects and those that have been attacked in service have never been fully characterized (as discussed in Appendix A). Systematic laboratory studies of the effects of heat treatment and stress could build confidence in the conclusions offered here and provide valuable information for life assessment and risk evaluation.


6

2.0 INTRODUCTION

High Temperature Hydrogen Attack (HTHA) considered here is the appearance of voids or cracks containing methane at grain boundaries and inclusions of some steels when they are exposed to hydrogen environments. It occurs in carbon and low alloy steels at temperatures above at least 400ºF because carbon and carbides in the steel may react with dissolved hydrogen to form the nondiffusible hydrocarbon gas. The rate of formation of methane is expected to depend on the temperature, amount of hydrogen dissolved in the steel, and many metallurgical factors, especially the thermodynamic activity and concentration of carbon in solution. HTHA was first reported about 75 years ago (7), but is not yet adequately understood. It causes concern and occasional failures in the refining, chemical and power industries. The purpose of this report is to offer a technical basis for the pressure-temperature operating limits provided by API RP 941 for equipment in petroleum industry hydrogen service. Conservative guidance in API RP 941 has been drawn mainly from reports of experience as described in Appendix A. In contrast, there has been little acceptance by industry of quantitative pressure-temperature attack limits developed from first principles by researchers nor has there been progress on remaining life estimation tools for possibly damaged steel. This report includes the positions of researchers in Japan, Europe, and America as recently communicated to an API RP 941 Task Group. Included herein is an overview of concepts that have been used over the years to understand the technical basis for operating limit curves of API RP 941. These concepts range from explanations based on extrapolation of simple exposure experiments to finite element models of the complex phenomena leading to void growth and cracking. Despite the efforts of dozens of researchers over the last 30 years, there has not been a successful and widely accepted synthesis of observations with the models for HTHA. The basic shapes of the curves can be justified, but their locations and the effects of applied stresses have not been adequately rationalized. Occasional problems with C – 0.5% Mo steel in long term service below the published limits for the alloy have raised concerns about the guidance found in API RP 941 with regard to other steels. For example, the question has been asked “Does resistance to HTHA decrease during time in service (as a result of aging)”. Before that question and others noted below can be answered, it is essential to see where we are regarding the fundamentals of hydrogen attack and the evolution of damage. Some further details and barriers to understanding behavior and future directions for needed studies are indicated in Appendix B and Appendix E. It should be noted at this point that the frequent references to C-0.5% Mo steel in this document should be viewed as applying to Mn-0.5% Mo steels and steels of similar composition. The current state of firm knowledge does not permit drawing fine distinctions with regard to elements that are not strong carbide formers. Based on this review of research, a semi-quantitative technical basis is offered for the current pressure-temperature curves and computing time to failure under the combined effects of hydrogen pressure and stress. These principles should be refined so they might be applied to handle excursions above the Nelson Curves as described in Appendix C and Appendix H (the latter prepared by the API Task Group).


7

The mounting body of evidence that the C - 0.5% Mo steel lines proposed by Nelson were inappropriate (Appendix A) stimulated interest in developing a technical basis for establishing recommended safe operating boundaries for carbon and low alloy steels in high temperature hydrogen service. Among the questions about why Nelson‟s curves for C - 0.5% Mo steel needed to be repeatedly changed were the following: (Note that, today, some engineers are asking the same questions about 1Cr – ½ Mo and 1-1/4Cr – 1/2 Mo- Si steels.) i) Were operating conditions reported to Nelson correct?

ii) Were the lines incorrectly drawn?

iii) Did the materials that displayed attack in service change in time so they became increasingly susceptible to attack over many years?

iv) Is there an incubation period after which attack is rapid?

v) Did attack proceed from the start of exposure, but methods of detection were too insensitive to disclose the progress of damage?

vi) What role is played by corrosion products or cladding in impeding hydrogen entry into the steel and subsequent hydrogen attack?

vii) How might one do a Fitness-for-Service evaluation?

To address these questions and more, an examination of public domain information and relevant research has been accomplished. Researchers in Japan and Europe were enlisted in a cooperative effort under the API program. Much of the discussion about published laboratory test results was accompanied by what seemed at the time to be very reasonable supposition and conjecture, but which has not stood the test of time. The utility of information from operating plants is not useful in many cases because of inadequate descriptions of the exposure conditions, materials studied and degree of attack. The author has tried to separate facts from conjecture. However, fully satisfactory answers to the when and where of attack cannot be provided because of the limited reliable information available, the complexity of the phenomenon and the many variables that play a role. What follows instead are current positions, some practical tools and suggestions for resolving the areas of ignorance. This is done in the hope that this work will form the basis of a better directed experimental program to get the additional information that engineers need.

2.1 What is Hydrogen Attack? API RP 941 recognizes two basic types of hydrogen attack:

1) Reactions on surfaces exposed to the process stream and evidenced by decarburization. 2) Internal damage as evidenced by voids (“bubbles”), cracks, (micro) fissures and blisters,

sometimes associated with internal decarburization. Both types of attack derive from the reaction of carbon or carbides with hydrogen to form methane. Methane formed at the surface escapes and the decarburization usually does no significant damage to the load carrying capability of the component. Methane formed internally is trapped since molecules of methane cannot diffuse through steel. This internal attack is of primary concern here.


8

Methane pressures at the site of internal attack can build to high levels forming or growing voids or cracks. The steel may then suffer progressive internal ruptures which reduce strength and toughness. With time, the depth at which methane appears may increase and the damage may propagate through the thickness, well beyond the surface exposed to the hydrogen. If undetected, cracks in stressed components may grow to failure. Weakened, fissured material may fail on startup due to loss of tensile strength or inadequate toughness. 106HFigure 1 to 107HFigure 5 indicate the key features of internal HTHA. A simple and thermodynamically correct way to view HTHA , even for steels with complex carbides is to consider the behavior of hydrogen and carbon in the steel (8-16). Then the reactions may be expressed as follows:

Initially [1]

The thermodynamic driving force for the methane reaction, the negative free energy change for [1], increasingly favors methane formation with decreasing temperature as shown in 108HFigure 6. As described below, calculations show that the internal pressure that may be produced by the trapped methane can exceed the nominal tensile or creep strength of the steel at the reaction temperature, even for low concentrations of carbon and hydrogen in the steel. Fortunately, the kinetics of the reaction slows with decreasing temperature and we expect an apparent threshold temperature for each steel, below which the attack rate is negligible. Because of the high availability of carbon in carbon steels and the carbon 0.5% Mo type alloy steels, reactions may proceed at significant rates even at low temperatures provided hydrogen pressure (and therefore solubility) is high. Materials rich in carbide-forming elements (low carbon concentration and activity) can be used at high temperatures and pressures. 109HFigure 7 shows how this consideration results in comparable reaction rates for the various classes of materials found in API RP 941 in their respective correct ranges of temperatures and pressures (110HFigure 8). Thus, it is concluded that, for each alloy considered in API RP 941, reaction kinetics govern the pressures at which attack is observed in the lower temperature regime and the pressure at any temperature depends on carbon availability. The free energy of methane formation for reaction [1] above is related to the pressures or activities of the participating species as follows:

[2]

Where the terms are defined as follows:

= thermodynamic effective “pressure” called the fugacity of methane. Methane is a very nonideal gas at the pressures of interest. Therefore, fugacity, not pressure, is used for the calculation. The actual pressure of methane formed may be much lower than the fugacity, but may be very high relative to the creep or even tensile strength of the steel.


9

= partial pressure of hydrogen. It is assumed hydrogen behaves as an ideal gas, so pressure may be used in the calculation. Also, for the calculation, it is assumed that the hydrogen pressure in the gas and the solubility in the solid are related through Sievert‟s Law, i.e. concentration is proportional to the square root of the partial pressure in the environment. Note the frequent references in this report and figures herein to hydrogen pressures or partial pressures in psi units should be taken to mean psia.

= activity of carbon in solid solution and in equilibrium with carbides in the metal. The value to use is for carbon dissolved in the iron matrix in contact with the least stable carbide phase in the steel. For alloys rich in carbide formers such as, molybdenum and chromium, a number for the activity much smaller than unity may be correct, say .01 to .1 when referred to graphite as carbon‟s standard state where activity is equal to 1. Where carbide formers are not plentiful and Fe3C, cementite, predominates, the effective activity (relative to graphite) may be viewed as greater than 1 since Fe3C is metastable (not an equilibrium phase) and numerically exceeds that for the graphite reaction.

The values of carbon activity in equilibrium with mixed (alloyed) carbides in low alloy steels have been calculated by many researchers. While they are probably accurate, the results are only approximately confirmed. This is because it is not convenient (or even understood how) to practically measure the activity of carbon in a solid alloy. Additionally, the actual activity of carbon in solution depends on heat treatment, temperature, and time. Equilibrium may not be reached and use of the equilibrium calculated value may not be appropriate. Unfortunately, the value of carbon activity chosen significantly affects the estimate of the most important driving force for HTHA, the methane pressure. If the least stable carbide is one of vanadium, molybdenum or chromium, the methane fugacity and potential pressure are greatly reduced as indicated by the above equation and discussed later.

= Free energy of formation for methane. If calculated on the basis of a reaction with cementite, the magnitude of the free energy change increases significantly over that calculated using graphite as carbon‟s standard state. The potential methane partial pressure can be very large.

Keq = Equilibrium constant

R and T have the usual meanings in any units consistent with those used for .

A brief discussion of the role of kinetics on HTHA may be helpful here. Among the reactions possible leading to the formation of methane in steel are the following, where (ad) suggests that the reaction takes place on a metal surface, external or in an internal void (cavity):

[3]

[4]


10

[5]

[6]

[7]

[8]

[9]

[10]

Allen, et al. (17) proposed that the rate of methane formation should be proportional to the fourth power of hydrogen content multiplied by the carbon content. This is reasonable from ordinary reaction rate theory, although a lower exponent for the concentration dependence is possible, as might be indicated by Equation 3 through Equation 10 above. Few details of the reactions in steel are known. Grabke and Martin (18) experimentally determined the kinetics, the rate of formation and decomposition of methane during decarburization at an iron surface. If one assumes that the basic processes are similar to what would happen in a void and if dnc/dt denotes the rate of change in the number of moles of carbon in the metal, Grabke‟s and Martin‟s results are:

[11]

where: A is the surface area of involved, K1 is the forward reaction constant for formation, Cm is the carbon concentration at the surface in moles-cm-3, K2 is the reverse reaction constant,

and are the fugacity and pressure of methane and hydrogen respectively in

atmospheres. The values of K1 and K2 measured by Grabke and Martin were:

[12]

[13]


11

These rate constants were obtained for pressures around atmospheric, but lacking other data, we assume that they can be applied at the pressures within voids as considered here.

There are many impediments to applying these equations but, since the intent here is to roughly identify the boundaries where reaction rates may govern, the magnitudes and trends indicated should be explored. The forward reaction, equation [12], is of particular interest. Logically, the reverse reaction [13] should not be of concern unless significant methane pressure has built up in the steel, so it need not be considered in determining when conditions are amenable to production of methane. There are three variables of concern here – temperature, pressure, and carbon content (related to carbon activity and the precipitated carbides). Calculations using the above equations show that the forward reaction starts to increase rapidly at high pressures in the neighborhood of the temperatures we see for hydrogen attack limits in API RP 941 for carbon steels, as shown in 111HFigure 8. Attack rates are equally rapid at low pressures only at high temperatures for steels with high carbon activity. In order to achieve sufficient reaction rate where carbon activity (dissolved carbon) is low, as in steel containing strong carbide forming alloying elements, the temperature and pressure must be high enough that sufficient hydrogen may be dissolved in order to compensate for the lack of availability of carbon. This concept does not need to deal yet with the number of sites at which the reaction can occur. Low energy interfaces, inclusions, coarse carbides, and other heterogeneities of the microstructure will influence relative susceptibility to attack. The point is that if the driving force and the rate are insufficient, attack will not occur. Above the minimum pressure-temperature limits, relative rates are of concern. Appendix C deals with some additional aspects of the rate question.

Finally, in support of this rate limitation concept, it is obvious that the methane reaction and/or cavity nucleation rates must be very slow at low temperatures given the potentially high methane pressures that could be generated given the increasingly large thermodynamic free energy change of the reaction with decreasing temperatures. Otherwise, attack would occur in common equipment operating at low temperatures, even at low hydrogen pressures

2.2 What is Fugacity and Why Is It Important

Methane does not behave as an ideal gas. As a result, the partial pressure of methane that can be developed in the steel is lower than the fugacity defined above. The fugacity-pressure relation of methane was reported about 25 years ago and is widely accepted by all investigators (8,19,20). It can be cumbersome to implement. For convenience, the approach adopted in the following text is based on an approximation that introduces very little error over the entire range of temperatures and hydrogen pressures of interest. 112HFigure 10A presents the relation used here - i.e.:

[14]

Strictly speaking, is a function of T and PCH4 (see Appendix F and Appendix G for a more lengthy discussion). However, when is an appropriately chosen constant, errors introduced in the calculated methane pressure, over the range of temperatures and pressures of interest, are small. They are also irrelevant, since we do not expect that equilibrium will be reached (nor would we know it , if it were) and there are many other, larger uncertainties in the calculations that must be made.


12

The equation shown above cannot be conveniently solved for when is specified

since the variable, , cannot be usefully mathematically separated in equation [14]. An

iterative solution must be used. When is a constant, the iteration is simplified.

The nonideal gas behavior of methane was not understood when many experiments were planned by early researchers (17,21-24). As a result, the data obtained could not be properly interpreted. This slowed progress in understanding HTHA. The equilibrium methane pressure increases with carbon activity. (See 113HFigure 10B.) The values of possible were always recognized to be of the magnitude of the tensile or creep

strengths of the materials of interest.

Caution: The units of and P must be consistent. A value of 0.005 for applies when

is in MPa and is usually the value shown in the literature. As noted above, the “constant”, 0.005, in the fugacity equation is not strictly correct for all methane pressures and temperatures. Changes of at very high methane pressures lead to noticeable differences in the fugacity estimates. The converse is not true. Estimates of methane pressure for a given fugacity are relatively insensitive to the value of used. A sensitivity study by the author over the range of fugacity values of interest showed the error in estimating methane pressure will not exceed about 5%. Therefore, 0.005 was used exclusively in this work.

The units of in the pre-exponential term in the above equation must be consistent with

those used for fugacity. Since, the equilibrium constant is calculated with the units of fugacity, atmospheres, methane pressure in the pre-exponential term must then be expressed in atmospheres when hydrogen pressure is also in atmospheres. Methane pressure is converted and expressed in ksi in this report for ease of discussing mechanical properties in the USA.

2.3 Details of the Attack Progression

2.3.1 Solubility Before discussing theories, data, and mechanisms, and attempting to predict of safe operating limits, a brief discussion of what occurs during void and crack formation may be helpful. On exposure to hydrogen at elevated temperature and pressure, hydrogen will be first adsorbed and then (absorbed) dissolved in the steel. (Equations [3] and [4]) The limiting (maximum) amount of hydrogen that can be dissolved in carbon and low alloy steels is calculated from an equation of the form:

[15]

where: A = a constant, and

= the hydrogen (partial) pressure, and

Q = an activation energy.


13

This is Sievert‟s Law. Typical values of the solubility may be seen in 114HFigure 11 based on a set of constants in 115HTable 1 used over many years in API publications. It should be noted that the amount of hydrogen dissolved in steel under attack conditions is very low, only a few parts per million (ppm) by weight, at most, as may be judged from 116HFigure 11 and 117HFigure 12. For carbon steel and C – 0.5% Mo steel, the dissolved hydrogen content under attack conditions appears to be on the order of 1 ppm by weight or less. ( 118HFigure 12) It is several times higher in low alloy steels where carbon activity is much lower. Lines of constant solubility in T-P space (119HFigure 12) bear remarkable similarity to some of Nelson‟s curves. They suggest that the attack limits may be related closely to the amount of hydrogen dissolved. Again, in the case of carbon and the 0.5% Mo steel, only about 1 ppm provides a sufficient driving force and rate at the temperatures of interest. Values of the constants A and Q should depend on alloy content, carbon content and heat treatment (microstructure). Data for the specific piece of steel used in a refinery application are never available. The solubility of interest for calculating the rate of hydrogen attack is the dissolved, atomic hydrogen content. Measurements of hydrogen solubility are rendered difficult by trapping of H2 molecules at microstructural defects, carbide interfaces and imperfections in steel. 120HTable 1 indicates the range of uncertainty when assigning values to hydrogen solubility in steel based on various data in the literature. Depending on how hydrogen content is measured some of it may be associated with trapped hydrogen molecules. A sound wrought steel product not previously exposed to a hydrogen atmosphere may contain trapped, molecular hydrogen exceeding 1 ppm by weight. The amount is a function of the melting practice, microstructure and processing. After exposure to a hydrogen containing environment, the amount trapped may be several ppm.

Table 1 –Solubility Ranges Calculated, Micromoles Per Cm^3 at Indicated Temperature, F reported for low alloy steels

Assuming a void volume in the material of 1x10-2 percent, (i.e. 99.99 theoretical density) and a trapped molecular hydrogen content of 1 ppm, we can calculate the possible pressure of hydrogen in voids as follows (assuming it does not dissociate and dissolve first): 1 ppm of hydrogen by weight is:

[16]

The volume of that mass at normal or standard conditions of temperature and pressure is:


14

[17]

The corresponding hydrogen pressure at 10-2 void percent would be in the ratio of the amount in [17] to 10-4 or on the order of 850 atm or 12,500 psi at ambient temperature. At operating temperatures for refinery equipment, the pressure of molecular hydrogen in the void would be about twice as high. The reaction to form methane would at first reduce the pressure (also some hydrogen would dissolve in the surrounding steel). Complete conversion to methane without further inflow of hydrogen would reduce the pressure by about a factor of two. At temperatures where the reaction rate is rapid and strength is low, trapping sites or voids might be natural HTHA nucleation sites. However, 1 ppm of hydrogen would consume only 3 ppm of carbon. The methane forming reaction must continue and use more than 100X as much carbon for swelling, void growth and eventual local internal decarburization. However, such voids or interfaces are reasonable sites for initiation of methane reactions. Where such nuclei are numerous, attack may be manifested relatively quickly and continue where sufficient hydrogen is available.

2.3.2 Incubation Times

The availability of easy nucleation sites is pertinent to the question of an incubation time for hydrogen attack. Investigators studying carbon and 2-1/4Cr–1Mo steels have shown that swelling occurs almost immediately on exposure to conditions favorable for hydrogen attack. The experiments of McKimpson and Shewmon (25), for example, disclosed immediate expansion. The widespread notion of an incubation period during which there is no attack may be attributed to the fact that in the early stages of attack voids are extremely small - only a fraction of a micron - and go undetected, using even the best optical microscopic techniques. API RP 941 refers to a period of no noticeable attack as the incubation period. Clearly, the duration of that period is dependent on the means of detection. As the years have passed, the means of detection have improved. Now, the original incubation curves should be viewed as the conditions for detection of attack in specific steels, rather than a threshold time for attack in an alloy grade. Given the variability of carbon and 0.5% Mo steels, it should be appreciated that some components may suffer severe attack or even failure in the time periods noted in API RP 941, while others might evidence little or no attack when studied over much longer periods of time, even using sensitive metallographic methods. For this review, McKimpson‟s (25) expansion rate data for carbon steels at various pressures and temperatures have been reanalyzed using the thermodynamically predicted equilibrium methane pressure as the driving force. This exercise suggested the usefulness of methane pressure as a key variable for predicting attack. (Many had suggested this in the past, but had not followed through due to uncertainty about the calculation or strength parameters). The

equation now found for McKimpson‟s data with an of .95 on the log of expansion rate was:

[18]


15

The low exponent (i.e. 2.268) for the methane pressure dependence is not inconsistent with current void growth models.( In Appendix G, a more complete discussion of values is provided and it would appear that the value should may depend on the alloy and temperature or even population and distribution of nuclei) McKimpson reported that after a period of relatively constant rate, the expansion rate increased exponentially. This is sensible. As voids grow, additional nuclei are generated in the immediate neighborhood because of the local stresses around the voids and the expansion rate accelerates.

2.3.3 Growth of Cavities

In the usual general approach, the growth rate of voids is divided into a volumetric growth rate

due to the diffusion of atoms into the grain boundary, , and a contribution resulting from

creep flow in the adjacent grains, :

[19]

The two contributions depend on the cavity geometry and the driving stresses for void growth. For HTHA, the pressure pm inside the cavity is usually set as

[20]

and it is the dominant driving force. More recently, investigators account for the influence of additional stresses resulting from applied loading. The volumetric growth rate due to diffusion is given by

[21]

with f = a function of cavity size and spacing

Here, is a temperature dependent diffusion parameter, defined by in

terms of the boundary diffusivity and the atomic volume . For further details see Appendix F and Appendix G. The expression [21] can account for the effects of surface energy

and the surface tension Ts. and is the surface angle describing the cavity as represented in 121HFigure 2.


16

The creep contribution is taken to describe the growth of a hole in an incompressible power-law creeping solid subjected to a uniaxial tensile stress, . The growth of a cavity by creep depends strongly on the stress triaxiality. In the case of purely hydrostatic internal methane plus hydrogen pressure pm (corrected for the surface tension Ts as shown below) the volumetric growth rate expression sometimes used is:

[22]

with

[23]

The relative contributions of diffusion and creep to void growth are weighted through a mathematical parameter with the dimension of length

[24]

Large values of Lm compared to, for instance, the cavity spacing b, indicate that growth is dominated by diffusion, while smaller values indicate increasing contributions by creep deformation. Most modelers assume that nucleation of the cavities takes place at the beginning of the HTHA process over a time that is much shorter than the total lifetime. Hence, the half cavity spacing b is taken as a constant. Given the hydrogen and methane pressure and the cavity size, one can compute the instantaneous growth rate of a cavity by substitution in the volumetric growth rate expressions. The complete attack evolution is obtained by time step integration until a critical value of a/b is reached at which time cavity coalescence takes place (sometimes taken to be a/b = 0.7). Appendix G presents an overview of the most recent modeling of the growth behavior. While this approach provides an elegant picture of behavior, it has not been quantitatively successful because of the large number of assumptions and terms needed for prediction. Additionally, there is scant data on how diffusion, creep, damage, and continuous nucleation interact in reality. The sensitivity to nuclei size, spacing and continuing nucleation is great. Thus, these are elegant models with acknowledged limited predictive capabilities.

2.3.4 Hydrogen Diffusion

The maximum hydrogen contents shown in 122HFigure 11 and 123HFigure 12 were based on reaching equilibrium solubility as calculated using published values of Sievert‟s law constants. Equilibrium with the gaseous or liquid phase in contact with the solid may not occur if diffusion away from the exposed surface is rapid or ingress is retarded. Rates of absorption from the process stream and diffusion away from the surface determine the actual hydrogen content. Assuming the hydrogen content at the remote surface is zero, the flux, F, is:

[25]


17

where: D = Diffusivity,

= hydrogen content of steel in contact with the gas phase and

t = thickness

and

[26]

The literature provides a wide ranges of values for diffusivity and associated activation energies. As shown in International Institute of Welding (IIW) surveys (26,27), 124HFigure 13 through 125HFigure 17, diffusivity of hydrogen in steel depends on microstructure, alloy content and temperature. High values of diffusivity suggest proportionally greater thickness at which theoretical, maximum concentrations may not be achieved for a given flux. The range of uncertainty in diffusivity estimates is even greater than for solubility. Where diffusivity is high or adsorption rate low or impeded by a surface coating, a thin-wall component may have a lower hydrogen content than predicted by Sievert‟s law and will be at greatly reduced risk. This is because as described earlier, both methane pressure and reaction rate vary with the hydrogen concentration raised to a power. The net result is virtual elimination of the possibility of attack where hydrogen content is suppressed. The most important example of this is the application of clad or overlay on the ferritic steel surface as described below.

2.3.5 Hydrogen Entry into the Steel – The Role of Clad or Overlay

Impeding ingress of hydrogen into the metal lowers the hydrogen concentration and lowers effective or equivalent hydrogen pressure calculated by substituting the actual concentration into Sievert‟s equation. Austenitic stainless cladding is well known as a hydrogen barrier (28,29). Ferritic stainless steels are also effective as illustrated in 126HFigure 18 and 127HFigure 19. Corrosion products may be similarly beneficial. The mathematics of dealing with corrosion products is not obvious, since thickness, hydrogen solubility and diffusivity are not measured or recognized as properties of the corrosion product layer. However, a corrosion product may be treated as an equivalent or added thickness of cladding. As described in Appendix D, a coefficient obtained from measurement of the flux change associated with a corrosion product can be used to account for the presence of the corrosion product and used to estimate the backing steel concentration and equivalent hydrogen pressure. These equations, similar in form and function to those used to relate the effective hydrogen pressure under cladding or overlay and flux and hydrogen concentration can be used to account for the effects of corrosion products in a range of situations. Appendix D provides the basis for equations for calculating effective pressures in the backing steel and a diagram helpful in understanding the terms used. The terms used are as follows;

= Hydrogen flux associated with clad or corrosion covered component, respectively.

= Hydrogen partial pressure in operating environment.

= Hydrogen effective pressure at the clad/overlay surface closest to environment.


18

= Hydrogen effective pressure at the backing steel interface closest to environment.

= Diffusivities of hydrogen in clad/overlay and backing steel, respectively.

= Thicknesses of clad/overlay and backing steel, respectively.

= Solubility of hydrogen in clad/overlay and backing steel at 1 psi

hydrogen partial pressure, respectively. Terms are dependent on temperature and expressed in appropriate units.

= Pressure reduction ratio at backing steel interface due to presence of clad/overlay - i.e. /

= Pressure reduction ratio at steel interface in the presence of corrosion product. i.e., / in corrosion situations

= Coefficient used to relate flux to pressure differential for corrosion product. Assumed independent of temperature and pressure.

Then, for clad material:

[27]

Where is the effective hydrogen pressure at the bond line and is the hydrogen pressure in the process stream.

Setting [28]

Where D, C, and W are diffusivity, concentration and thickness and subscripts. B and C refer to backing steel and clad/overlay, respectively.

For clad with corrosion product (say sulfides) on the clad surface

[29]

It can be seen from 128HFigure 20 and 129HFigure 21 that the benefit (reduction effective hydrogen pressure) of a corrosion product and/or cladding (overlay) is dependent on the relative thickness of overlay and backing steel and temperature.


19

For ferritic (400 series, 9-12 chrome), clad or overlay, the relevant product of diffusivity and solubility is reported to decrease with increasing chromium contents up to about 15% by about a factor of 4 compared with carbon steel (30). This compares favorably with austenitic stainless steels where the product is usually taken to be about 10. This indicates that high-chromium ferritic clad/overlay can impart appreciable benefits to low alloy steels now in refinery service. The reductions in hydrogen pressure associated with a broad range of practical clad/ or overlay/backing steel thickness combinations are shown in 130HFigure 20 and 131HFigure 21. It can be seen that for thin components, the effect is dramatic and would explain optimism about C – 0.5% Mo steel if the experience with clad components is included with the input data in Nelson‟s plots without adjusting for the reduction in hydrogen pressure in the backing steel interface.


20

3.0 ATTACK OF MATERIALS CONTAINING Fe3C

Carbon steels and C – 0.5% Mo steels, may contain significant amounts of cementite (Fe3C or M3C in low alloy steels). Every metallurgist is taught that cementite is not thermodynamically stable relative to its components, iron and carbon (graphite), i.e. the free energy of formation of cementite is greater than zero. The consequence of that fact with regard to HTHA is that the pressure resulting from methane formation in the vicinity of cementite is higher than if carbon were in a solid solution containing more stable carbides or even graphite. Another way to look at the situation is that the activity of carbon in ferrite with cementite may be viewed as though it is greater than one, in comparison to the free energy change based on methane from graphite. Calculating the methane pressure using the free energy of formation from pure cementite, a carbon activity of one is appropriate. Alloying of cementite with molybdenum, chromium or other slightly soluble carbide formers reduces the free energy of formation of methane and carbon activity can be reduced accordingly, depending on the degree of enrichment with the elements cited. 132HFigure 22 shows the difference in the free energy change for methane from graphite and from cementite. Using the free energy of formation for the metastable Fe3C phase, the equilibrium methane pressure calculated in accord with equation [2] is higher than when the calculation is based on the reaction of carbon (graphite) dissolved in iron. Methane pressures for hydrogen in equilibrium with pure Fe3C at temperatures of interest here are shown in 133HFigure 23. From 134HFigure 23, it is obvious that at the equilibrium methane pressure calculated can exceed the tensile strengths of carbon and C – 0.5% Mo steels steel at these potential operating conditions, even at low hydrogen pressures. The fact that steel is commonly used with hydrogen at lower temperatures must mean equilibrium is not achieved - i.e. nucleation of cavities does not occur and/or the rate of the methane forming reaction is negligibly slow. As a result high-pressure equipment can be successfully operated at temperatures below those shown in 135HFigure 23. Because of the nonideal gas (fugacity relationship) for the behavior of methane described earlier, at low temperatures the equilibrium methane pressure in contact with cementite is relatively insensitive to hydrogen pressure (136HFigure 23). However, this insensitivity does not explain the flatness of Nelson‟s P/T curves at low temperatures with increasing pressure. The potential methane pressures at low temperatures are so unrealistically high, they can never occur. Failure would be expected at even low hydrogen partial pressures. Instead, the explanation involving kinetics offered earlier is the most plausible and consistent for the flat disposition of the low temperature portions of each of Nelson‟s curves indicating tolerance for very high hydrogen pressures below a threshold temperature range. In regard to the steeply rising portions of the curves, investigators initially speculated that methane pressure was the driving force for cracking, but they had no way of quantifying the pressure and time dependencies of attack. Now, methane pressure can be shown to correlate with the time dependent (isochronous) carbon steel and other attack lines shown in API RP 941 (137HFigure 24 and 138HFigure 25). While not describing any particular material, the carbon steel lines may be valid indicators of the correct trends of laboratory data for the materials reported to Nelson. A plot of reciprocal temperature and methane pressure for each of the times is shown in 139HFigure 25. The calculations upon which theses curves were based used the free energy of methane formation from unalloyed Fe3C. The standard state for each reactant in the equation was set at unit activity (H2 and Fe3C) to calculate methane fugacity and then its pressure.


21

The roughly linear dependence with methane pressure observed over the range of temperatures of interest suggests a possible important role of methane pressure in the kinetics of this attack. It appears to confirm a functional relation between time and methane pressure, especially at higher temperature where kinetics would be expected to be limiting. The behavior shown in 140HFigure 25 is approximately in the form:

. [30]

While Nelson‟s time dependent curves are clearly not in the correct positions for all carbon steels, the potential for correlating behavior with methane pressure is indicated. It is reiterated here that, at low temperatures, kinetic factors prevent achieving the high methane pressures calculated even at defects or inclusions where nucleation of voids is easy At high temperatures, equilibrium calculated pressures will not be achieved if surface connected cracks form. Thus, 141HFigure 25 and the comments here about the role of high temperature strength are presented to suggest behavioral limits and not the actual pressures achieved. Also there are substantial uncertainties with regard to the free energy calculations upon which the methane pressures are based. It has been reported that M3C in alloy steels contains small percentages of Mn, Ni, and carbide formers such as Mo or Cr that reduce the free energy of the cementite phase and therefore decrease the potential methane pressure. The recent reports about attack of C – 0.5% Mo steel below Nelson‟s curves suggests that such reductions in the free energy of Fe3C, even associated with molybdenum, are not large. The superior tensile and creep strengths of the 0.5% Mo and Mn-Moly steels may explain the original basis for optimism about the behavior of the low molybdenum steels. Geiger and Angeles (22) examined the equilibrium constant, K or Keq ( 142HTable 2) for the methane reaction, considering the effects of alloying elements on the free energy of formation. Their results are reproduced below. The expected alloy element content in Fe3C would only reduce K by at most 50% and would not have a drastic effect on the methane pressure, as may be judged from 143HFigure 10B. For example, comparing carbon activity of 1 and 0.5, the reduction in methane pressure is about 20%. Significant reductions in methane pressure would only be associated with the K values shown for carbides in 1 Cr-1/2 Mo steel or 2 1/4 Cr - 1 Mo steel with higher alloy carbides such as Cr7C3 or Cr23C6.


22

Table 2 – Values of Equilibrium Constants, K [Equation (2)], for Various Carbides

Carbide Values of K at Different Temperature

900oF 950oF 1000oF 1100oF

Iron Carbides

(1) Fe2, C( -carbide) 5.37x104 2.04x104 8.70x104 1.74x103

(2) Fe3C 11.00 5.65 3.10 0.98 Untempered Alloy Carbides

(3) (0.005Mo,0.995Fe)3C 9.26 4.79 2.67 0.910

(4) (0.01Cr, 0.005Mo, 0.985Fe)3C 8.49 4.38 2.48 0.846

(5) (0.0125Cr or Mn, 0.005Mo, 0.9825Fe)3C 7.69 4.00 2.28 0.771

(6) (0.0225Cr, 0.01Mo, 0.9675Fe)3C 6.85 3.60 2.03 0.699

(7) (0.03Cr, 0.005Mo, 0.965Fe)3C 6.65 3.50 1.98 0.680

(8) (0.03Cr, 0.01Mo, 0.98Fe)3C 6.09 3.31 1.67 0.649 Tempered Alloy Carbides

(9) (0.04Mo, 0.96Fe)3C in 1/2Mo Steel 6.93 3.62 1.99 0.708

(10) (0.04Mo, 0.96Fe)3C* 5.71 3.00 1.67 0.565

(11) (0.2Cr, 0.8Fe)3C in 1Cr-1/2Mo Steel 2.24 1.21 0.704 0.270

(12) (0.2Cr, 0.6Fe)3C in 2-1/4Cr-1Mo Steel 1.81 0.980 0.589 0.230

(13) (0.2Cr, 0.6Fe)3C 1.29 0.751 0.456 0.152 Pure Carbides

(14) Mo2C in 1/2Mo Steel 1.32 1.07 0.87 1.11

(15) Mo2C* 5.04x10-4 4.07x10-4 3.39x10-4 2.29x10-4

(16) Cr7C3 in 1Cr-1/2Mo Steel 1.33x10-2 1.12x10-2 9.85x10-3 1.16x10-3

(17) Cr7C3 in 2-1/4Cr-1Mo Steel 2.62x10-3 2.20x10-3 1.96x10-3 2.91x10-3

(18) Cr7C3 1.03x10-4 8.90x10-5 7.76x10-5 5.75x10-5

(19) Cr23C6 in 1Cr-1/2Mo Steel 1.01x10-1 9.15x10-2 8.44x10-2 1.09x10-2

(20) Cr23C6 in 2-1/4Cr-1Mo Steel 5.25x10-3 4.73x10-3 4.32x10-3 1.14x10-2 *Carbide is assumed to be in equilibrium with ferrite whose composition is the same as the metal content of the carbide.

Geiger‟s computations (22) for more complex carbides include assumptions about the ferrite content that are not always practical.


23

4.0 METHANE PRESSURE RELATIONS WITH CEMENTITE

Very high methane pressures are expected in the presence of cementite, even at low hydrogen partial pressures. At 450ºF to 550oF (144HFigure 26 to 145HFigure 28), at unit activity, methane pressures on the order of the true strength of ordinary steels would be achieved at less than 100 psi hydrogen. Alloying of M3C with small amounts of strong carbide formers like molybdenum, thereby reducing the potential carbon activity, only increases the limiting hydrogen pressure by 20% or less. In accord with Geiger‟s (22) work maximum reductions in activity by a factor of 2 to 4 are considered to be reasonable for purposes of understanding the behavior of cementite. At 600ºF (146HFigure 29), hydrogen pressures in line with service experience with carbon steel can be justified. As shown in the figure, 100 psi corresponds to methane pressures in the normal tensile strength range. Higher strength steels of the manganese and half-molybdenum types should be beneficial at 600ºF to 650oF (147HFigure 29 and 148HFigure 30). At the lowest activities expected for C - 0.5% Mo steels, hydrogen pressures above 400 psi to 500 psi would be on the margin. For many carbon steels, creep is anticipated at 650ºF and above. For them, only 100 psi hydrogen would produce methane pressures high enough to cause creep on a local scale and, as a consequence, void growth. Enrichment of the M3C with alloying elements would produce a stronger alloy and lower carbon activity. Still less than 200 psi hydrogen would lead to failure unless cavity nucleation and the methane reaction are slow. At 700ºF (149HFigure 31), all carbon and carbon-0.5% Mo type steels containing cementite would be expected to be in their creep ranges at the stresses of interest. Methane pressures in the neighborhood of only 30 ksi might lead to failure. Hydrogen partial pressures in the range of 80 psi to 150 psi would seem to be the maximum acceptable. As a point of reference at ASME design stresses (31) in the creep range, strain rate is usually negligibly small - less than 10-8 /hr. An internal methane pressure numerically equal to the design stress level would not be expected to be adequate to promote expansion of a void at a rate of practical concern.

Hydrogen partial pressures similar to those at 700ºF would seem to be limiting at 750ºF (150HFigure 32). Creep strengths in the range of 20 ksi to 25 ksi are the practical limit for these steels. Increasing the temperature to 800ºF reduces the stress above which creep rates would be acceptable by about one third, but only slightly increases the acceptable hydrogen pressure (151HFigure 33). The story is identical at 850ºF (152HFigure 34) where about the same hydrogen pressures correspond to the design creep strength at that temperature. At 900ºF, ASME design creep stress for carbon steel is about 6.5 ksi. A methane pressure of that value corresponds to the same hydrogen pressure shown in the above noted figures for lower temperatures - i.e. about 100 psi at the higher values of activities and at most 200 psi for M3C with the lowest activity of carbon (i.e. with the highest percent alloy content). Finally, at 1000ºF, where the ASME design value for carbon steel is 2.5 ksi, somewhat higher hydrogen pressures might be tolerable (153HFigure 36).


24

It is remarkable that throughout the creep range, the strength reductions for these cementite containing steels are almost exactly balanced by changes in the free energy-fugacity relations to give nearly constant acceptable hydrogen pressures. There is no fundamental reason why the free energy of formation of methane and the creep strengths of steels should engage in this balancing act. It is amazing that Nelson suggested the consequences of this interaction in his predictions.


25

5.0 STEELS WITH MORE STABLE CARBIDES

Steels containing elements with a strong affinity for carbon (i.e. molybdenum, chromium, vanadium, titanium, etc.) are strengthened to resist creep deformation. These elements also decrease the methane pressure possible through reductions of the carbon activity and of the reaction rate. The amount of carbon left in solution correlates with activity as shown in 154HFigure 37. Thus the reduction is beneficial both kinetically and thermodynamically. If large improvements in resistance to HTHA are to be obtained there must be sufficient alloy content to substantially eliminate all of the metastable Fe3C or similar M3C carbides,. It is the least stable carbide in a material that governs its resistance to hydrogen attack.. The trend of the effect of carbon activity on methane pressure may be judged from 155HFigure 38. For this report, we can use rough estimates of dissolved carbon content as the activity in the more highly alloyed material, rather than complex calculations of the free energy of the carbides. Both routes should lead to the same calculated methane pressure because carbides in the steel are assumed to be in thermodynamic equilibrium with the carbon in solution. The more stable the carbide, the lower the amount of carbon left in solution and the lower the methane pressure. A note of caution is offered here. A detailed examination of the results of calculations of the free energy changes associated with complex carbides shows the corresponding methane pressures and carbon activities to be extremely sensitive to assumptions made about the resulting compositions of the adjacent solid solutions. In reality, PWHT times and temperatures and cooling rates influence the alloy content of the carbides and ferrite. The result is substantial uncertainty about the limits on the behavior to be expected. The equilibrium equation for formation of methane shows that the fugacity of methane is proportional to the carbon activity. Lower concentrations of dissolved carbon also slow the reaction producing methane. The presence of strong carbide formers then provides three-fold benefits - strengthening, methane pressure reduction, and slower methane formation.


26

6.0 BEHAVIOR OF STEELS WITH COMPLEX ALLOY CARBIDES

Included in this group are alloys ranging from 1-1/4Cr–1/2Mo to higher chromium-moly-vanadium steels. The addition of elements with strong carbide-forming tendencies reduces the activity of carbon in solution and increases tensile and creep strengths of the steel. First, we will look at the potential effects on activity and temperature. 156HFigure 39 shows the behavior for carbon activities from 0.01 to 0.15 at 450ºF. At this low temperature, it is unlikely that the methane forming reaction would proceed at a significant rate. It is also unlikely that the nucleation rate would be significant in modern, clean, higher-strength steels. Thus, almost any pressure can be tolerated. 157HFigure 40 suggests that at 700ºF, steels with activities below about 0.06 can endure very high hydrogen pressures. Tensile strengths should not be too low and creep rates would not be of concern for steels richer in alloy content than 1 or 1–1/4Cr–1/2Mo steels. The expected activity levels of 0.15 or 0.10 for those materials would be satisfactory, even tempered to relatively low tensile strengths. For 1 or 1–1/4Cr–1/2 Mo steels, it appears that pressures of the order of 1200 psi to 1500 psi might be acceptable, in line with experience with that alloy. For activities expected with 2-1/4Cr–1Mo steel, say in the neighborhood of 0.06 or lower, acceptable hydrogen pressures would be higher, perhaps above 3000 psi. The reaction rate would be slow given the very low concentration of free carbon and the low temperature. At 750ºF (158HFigure 41), acceptable hydrogen pressures increase, but the low chromium steels would be questionable for service if carbon content is high and alloy content is insufficient to assure freedom from M3C type carbides. More than 1000 psi hydrogen might be acceptable for the leanest alloys. The threshold for creep may be at about the yield strength levels for these steels at 800ºF (159HFigure 42). Kinetics of the methane reaction might be slow given the low amount of dissolved carbon in the highly alloyed materials. However, for 1 or 1-1/4 Cr alloy steels (activity 0.1 to 0.15) with yield strengths at temperature in the 30 ksi to 40 ksi range and for 2-1/4Cr–1Mo steels (activity 0.03 to 0.06) with 5 ksi to 10 ksi higher yield strengths, the acceptable hydrogen pressures seen in 160HFigure 42 are in line with common applications for the respective alloys. 161HFigure 43 to 162HFigure 45 depict potential methane pressures in the critical range of temperatures from 825ºF to 900ºF. These temperatures are of interest in predicting the behavior for high temperature hydroprocessors. Over this range of temperatures the reductions in creep strength with temperature are not exactly offset by the changes in the free energy of methane formation and the resulting reductions in methane pressure. However, the corresponding decreases in limiting hydrogen pressures with increasing temperature are small over this range. With this picture of the importance of both strength and reaction kinetics as influenced by carbide precipitation, it may be instructive to briefly consider past research on hydrogen attack phenomena before proposing details of a technical basis for arriving at safe limit curves.


27

7.0 RESEARCH STUDIES

The API RP 941 document covers several low alloy and carbon steels produced in many product forms and for many specifications. Broadly speaking, these are carbon steel, C - 0.5% Mo steel, 1Cr – and 1-1/4Cr – 1/2Mo steel, and 2-1/4Cr – and 3Cr – 1Mo steel. However, most studies have been devoted to C – 0.5% Mo steel and 2-1/4Cr–1Mo steel. As a result, the observations herein are skewed by the behavior reported for these two very different steels. Shortly after adoption of API RP 941, considerable research was funded by the petroleum industry. Studies by Allen (17,21), Geiger (22), Vitovec (24), and others were creative, thoughtful, and informative. However, they were limited in their ability to interpret results because of their lack of understanding of the nonideal gas behavior of methane and the crude metallographic tools to study void growth available at the time. Then, starting about 25 years ago and for a period of about 10 years, there was intensive experimental and analytical work devoted to explaining HTHA. It was aimed at predicting the proper positions of the Nelson curves. Key university investigators were Johnson (10,32), Shewmon (32-46), Li (47-49), and Odette (8,11). They proposed models that continue to be elaborated and refined today with advanced computational tools, but without major changes in the approaches. Extensive experimental work from Japan, focusing on C - 0.5% Mo steel, was reported in the same time frame (50-60). Except for MPC‟s test programs on the effects of stress on HTHA of 2-1/4Cr–1Mo steel (61-64), experimental work was abandoned in the 1990‟s. Recently, parallel programs (9,12-16,65) in Japan, Europe, and America have addressed HTHA, but primarily from the analytical rather than the experimental side. The studies have not been entirely complementary because of the differing approaches taken. However, the international activities bring the earlier concepts up-to-date and provide some new insights (See Appendix F and Appendix G). A lengthy, scientific treatise on HTHA research is avoided here because it would be inconvenient for a petroleum engineer to put into a practical context, would submerge the important points in scientific jargon, and would be repetitive of the excellent literature cited above. Interpretation and application are emphasized here at the expense of scientific and mathematical detail.


28

8.0 STUDIES

The post-Nelson studies of HTHA referenced above may be divided into several categories by date and investigator or program. These were:

1) Numerous API funded studies, principally by Allen, Geiger, Vitovec and their colleagues in

the 1960‟s. 2) Intensive research at universities with U.S.D.O.E. support in the 1970‟s and early 1980‟s. 3) MPC‟s experimental studies of the effects of hydrogen and stress on high-strength chrome-

moly steels in the 1980‟s and 1990‟s. 4) Studies of C – 0.5 Mo Steel in Japan in the 1980‟s and 1990‟s under JPVRC. 5) Modeling of void growth and decarburization in chrome-moly steels using FEA in Europe in

the late 1990‟s and beyond. 6) API / JPI studies of C – 0.5% Mo steel in Japan and America in the 1990‟s to date.

API took the initiative to understand HTHA after Nelson‟s work. Their committee-directed studies provided insight into the effects of some variables on attack rates, but at pressures and temperatures far too high to represent service. These studies included examining the effects of temperature, cold-work, and composition (17,21). However, the tools of the day for modeling behavior were limited and the details of the attack process were inadequately understood. It is difficult to use that early data on the effects of temperature, pressure and metallurgical variables to verify new models because the degree of attack was not quantified and the materials tested were not adequately described or characterized. Enlightening work on attack rates was performed by Steuber and Geiger (66), and later by Geiger with others (22,67). In their work, the progress of attack in pressurized tubes was monitored by methane release rates, metallographic examinations and hydrogen flux measurements. These are the few studies in which the kinetics of attack were monitored and from which attack rates might be deduced. They exposed the interiors of carbon and low alloy steel tubes to high pressure, but relatively too high-temperature hydrogen for practical extrapolation purposes. Unfortunately, their analysis of their data may have included errors or oversights that led to some confusing or incorrect conclusions. Also, they were hampered by the shortcomings of metallurgical and thermodynamic tools of their day. Much more could be learned today with new experiments with modified test methods and analytical approach for the data. Even so, because of the importance of their observations, it is useful to consider their results in some detail here.

Steuber and Geiger observed that permeation rates through the tube walls increased with time because surface connected fissures formed by attack reduced the effective wall thickness of the tubes. Fissuring was detected when methane was released from the tube‟s interior wall surface. They concluded that leveling off of the hydrogen permeation rates with time after an initial period of linear increase could be related to a threshold methane pressure (fugacity) at the deepest point of attack and cessation attack at that depth. One problem is their calculation of the methane pressure (actually fugacity) at the deepest point of observable damage may not have adequately adjusted for surface connected fissuring. They acknowledge the surface connectivity of fissuring in their discussion of peaks in methane measurements and their data appear to be accurate and consistent. However, reanalysis of their permeation data today suggests that only 40% to 45% of the total depth of attack was associated with surface connected fissures. A correction for this error would provide an estimate of the hydrogen concentration (methane pressure) at which damage for the period of exposure could no longer be observed metallographically. Modifying and repeating their experiments today might be fruitful.


29

A second problem with their conclusions is the leveling off of permeation and the apparent reduction in the rate of attack reported by Steuber and Geiger was due to a limitation of the rate of hydrogen absorption into the tube wall, not an attack threshold. The uncracked wall thickness of their tube wall decreased with time. When the wall is too thin, the hydrogen content in the metal cannot reach Sievert‟s solubility limit unless the rate of adsorption at the solid/gas interface can increase in proportion to the reciprocal of the remaining ligament. The maximum concentration of hydrogen during Steuber‟s and Geiger‟s (single-side) exposure of pressurized tubular specimens was probably limited by the absorption rate of hydrogen from the pressurized tube interiors. In some cases the wall thickness was reduced only a small amount before attack leveled off, but because of the high hydrogen diffusivity at the very high temperatures they used, it is possible that the flux (absorption rate) limited attack in their thin wall experiments. When adsorption limits the flux, the concentration of hydrogen in the metal may be far from the equilibrium value. Since methane fugacity varies with the square of the effective hydrogen pressure and the hydrogen pressure varies with the square of concentration, small reductions in surface hydrogen content have large (fourth power) effects on the actual methane pressures that would be achieved. The methane fugacity-pressure (nonideal gas) relationship and access to computing capability enabled Odette (19,20), Johnson (10), Li (48) and others to seriously attempt to model behavior. These researchers used essentially the same creep/diffusion void growth models, thermodynamic relations and nonideal gas equations described earlier. However, their work was limited by lack of knowledge of key materials properties and behavioral data. These deficiencies included data on materials strength, carbon activity, surface and grain boundary diffusivities, bubble growth kinetics and the effects of temperature and impurities on each of the above. These researchers assumed “Nelson Curves” to be correct for the purpose of checking their models (as is done herein). However, in most cases the curves appeared to be too conservative in light of experience and there was no validation by long-term tests. Therefore, the curves lacked credibility and were never accepted. Reports of successful experience with chrome-moly steels, beyond the limits suggested by the researchers, led to the conclusion that the models might be missing something. Interest in using those analytically derived results eventually declined. JPVRC studies (51,57,58,68-71) of C - 0.5% Mo steel revealed significant differences between two heats that were extensively studied. As expected, there were beneficial effects of tempering and adverse effects of very high temperature weld thermal cycles on susceptibility. It was concluded that Shewmon‟s and (Appendix F) equations for the attack rate might be used to describe time – pressure – temperature relations of the general shape of Nelson curves for C – 0.5% Mo steel. The coefficients in the and curves for one steel studied were adjusted to be close to the final C – 0.5% Mo line in the then current edition of API RP 941. This basic or curves can be shifted up or down in P-T space by adjusting one or more coefficients to intersect limited data for any material.


30

Hattori (73,74) modified the concept by including a factor for “bad” M23C6 carbides in the 0.5 5 Mo steel. However, others have not corroborated his conclusions about M23C6 or concurred about its existence in commercially supplied, normalized and tempered heats of C-0.5% Mo steel. Details of the carbide Hattori reported went unconfirmed by others and the HAT or Hydrogen Attack Tendency Chart proposed by Hattori is not in widespread use. However, his work did put the spotlight on the detrimental effects of coarse carbides. All agree that coarse carbides may serve as plentiful nucleation sites and should be viewed negatively. Other JPVRC studies in Japan consisted mainly of exposures to very high pressure hydrogen for short times followed by measurement of tensile or impact properties. That work did not include modeling and did not incorporate predictions based on materials‟ properties, thermodynamics, or reaction rate measurements to clarify or quantify details of behavior. However, the experiments by JPVRC did contribute to the growing body of evidence that the early C – 0.5% Mo steel lines had been incorrectly (optimistically) placed. High impurity content was concluded to be an indicator of increased susceptibility to hydrogen attack. The material found to be inferior in those studies had higher impurity (S, P, etc.) content. (It also had high aluminum content.) Unfavorable experience in the field with heats containing high levels of tramp elements seems to validate the conclusions of JPVRC.

Shewmon‟s and Pw concepts were reexamined in the JPVRC study and during the work of the Japanese team in API‟s activity, as described in Appendix F. The discussion in Appendix F uses the C – 0.5% Mo lines in API RP 941 as the template for projecting behavior to other alloys. Acknowledging the importance of methane pressure the investigators rationalized the curve shape without getting into rigorous detail about the attack process. They modified Shewmon‟s equation for the initial attack (expansion) rate to predict failure time and quantified the expected trends based on temperature, hydrogen pressure and fugacity of methane. A comparison of and is worthwhile at this point. They are not equivalent or interchangeable. Setting t equal to life

[31]

or

[32]

while

[33]

or

[34]

In the case,


31

[35]

while in the case

[36]

predicts a much stronger dependence on hydrogen pressure (hence hydrogen

concentration) than does . This is a big difference. At temperatures of interest, for C – 0.5%

Mo steel at 700ºF, the pressure dependence of is fifth order, a surprisingly high number in light of models and observations of gradients in through wall damage. Actually, neither Pv or Pw does well predicting observed damage gradients. The through-wall damage gradients implied by and by at usual operating temperatures for C – 0.5% Mo are shown in 163HFigure 47.

The gradients depicted in 164HFigure 47 would be expected prior to the appearance surface connected cracking or blistering. It is assumed Sievert‟s and Fick‟s Laws apply - i.e. the effective hydrogen pressure varies with the square of the concentration of hydrogen at any point in the metal and flux is proportional to the (linear) concentration gradient. Additionally, it is presupposed that the delivery (diffusion) of carbon and hydrogen to sites (voids) for the reaction to form methane does not limit the rate of attack. The result is that damage is predicted to decrease with the through-wall distance more rapidly than has typically been observed. Examination of the attack gradients in components exposed to hydrogen on one side suggests that gradients shown in 165HFigure 47 are much too steep. For comparison, a profile shown in the 166HFigure 47 uses methane pressure as a driving force. It suggests a flatter curve and deeper penetration of visual damage. Attack is often seen more than halfway through the wall in one-sided exposures. Such observed behavior cannot be reconciled with expressions of the and

form. Also, calculations for and do not take into explicit account for acceleration of the attack with the passage of time due to the simplifications introduced to make the model more practical for implementation. Appendix G provides insight into the complexity required to fully model all the factors contributing to the time dependence of damage. .


32

9.0 EFFECT OF STRESS

Few studies of the effects of stress on hydrogen attack have been reported. The most comprehensive program was that of MPC. MPC studies of the effects of stress have been reported in detail elsewhere (64). They showed that for stress rupture specimens tested in hydrogen under moderate applied stresses, HTHA was accelerated and occurred below Nelson‟s lines. The materials studied were base metal and welded 2-1/4Cr–1Mo and 3Cr–1Mo steel tempered to typical as well as relatively high strength levels. Some observations previously reported are updated here. The first MPC studies involved base metal which showed attack, in long-term exposures. Then, the work explored the behavior of weldments. The heats studied were representative of modern low impurity materials. The weldment from which specimens were made was a typical submerged arc joint between two heavy plates from different producers. The two heats of 2-1/4Cr–1Mo utilized will be referred to herein as the 12-inch (30-cm) and the 13-inch (32.5-cm) heats. While carbon activity was not measured directly, there was an expectation that for the heat treatments studied reduced carbon activity should be associated with increased tempering temperature (or Larson-Miller parameter). Values of 0.03 and 0.06 are used elsewhere in this report for carbon activities resulting from high (1275oF) and low (1165oF) temperature tempering, respectively. These are estimates and are consistent with measurements made by Shewmon and referenced in another MPC study, (75,76) For the tests in hydrogen, welded samples were heat treated in two ways. One heat treatment provided an accumulated temper parameter of 36.79 equal to 16 hours at 1275oF. The other parameter value was 34.74 for 24 hours at 1165oF. These heat treatments corresponded to tensile strengths of approximately 85 and 102 ksi, respectively, and cover the strength range found for accelerated cooled materials placed in service over the last thirty years. A large database compiled from creep-rupture tests of these same materials and weldments in air was used for comparative purposes.

The effect of stress may be seen directly from several tests of the 12-inch weldment shown in 167HFigure 48. For material given the 1275oF temper, three tests at 850oF show a relatively small increase in life (a factor of 2) with decreasing stress between 37 and 28 ksi. In air, an increase in life by a factor of 10 would be expected. Further reduction of stress to 25 ksi suggests a flattening of the slope, but the relative life fraction as compared to tests in air was not improved. At 825ºF, the slope of the rupture line was about the same as in air, but lives were only about one-third as long. At 900ºF, the effect of hydrogen was greater and the curve steeper than in air. The locations of the failures were in the heat affected zones of the weldments and not in the base metal or filler metals. For the same material tempered at 1165ºF, tests at 3000 psi hydrogen and 900ºF showed only a small effect of stress at 28 ksi and 25 ksi ( 168HFigure 48), but a greater increase in life with further stress reduction. This data suggests a complex interaction between hydrogen pressure and stress as described later.


33

10.0 EFFECT OF TEMPERATURE

As shown in 169HFigure 49, for the 12-inch plate tempered at 1275ºF, there was only a factor of 2 increase in life when the test temperature was reduced from 850ºF to 825ºF at 37 ksi. A larger increase in life at 33 ksi was observed. This may be only scatter in data considering the life at 30 ksi was shorter. The Larson-Miller parameter for behavior in air leads to the expectation of a factor of about 3 for a 25ºF change in temperature. At 28 ksi, decreasing temperature from 900ºF to 850ºF resulted in an increase in life by only a factor of 3 - i.e., again less than a factor of 2 per 25ºF instead of 3. However, at 25 ksi, the ratio was somewhat greater. Clearly, under hydrogen, the effect of temperature is dependent upon stress and may be complex in the vicinity of the Nelson Curve limits. This is to be expected since decreasing temperature increases the potential methane pressure relative to the applied stress and should shorten life. While, creep rates are slowed, details of the interaction of applied stresses and internal methane pressures is a matter for speculation at this point.


34

11.0 EFFECT OF HYDROGEN PRESSURE

The effect of hydrogen pressure was observed to be remarkably consistent and apparently independent of material. The trend may be seen in 170HFigure 50. All data shown are for material tempered at 1165ºF. Starting with the tests on 12-inch plate at 850ºF and 33 ksi, life increased by a factor of 2 from 10,387 hours to 22,363 hours as hydrogen pressure decreased from 3000 psi to 2000 psi. At 900ºF and 28 ksi, the relative effect was precisely the same, life increased from 3,697 hours to 7,490 hours with the reduction in pressure. For the tests of the weldment involving the 13-inch plate, the life observed increased from 15,327 hours to 31,450 hours. However, the failure location changed from fine grain to coarse grain heat affected zones. Of course, there was also extensive hydrogen attack damage in the fine grain region at failure. It was to be expected that rupture life would be an inverse function of pressure. The data collected suggest an exponent of about 2 - i.e., life varies with the reciprocal of pressure squared. Shewmon (35) projected approximately that figure for quenched and tempered material. He stated that strain rate due to bubble formation should be proportional to hydrogen pressure to the 2.2 power.

11.1 Materials Issues

Tests in air had revealed that the 12-inch material, welded or not, always exhibited superior behavior as compared to the 13-inch material. The behavior of the two weldments in hydrogen can be compared at each of five identical test conditions studied ( 171HFigure 51). For materials heat treated at 1275ºF, the tests at 850ºF, 3 ksi hydrogen pressure and 33 ksi stress yielded 5,782 hours for the 12-inch weldment and 22,192 hours for the 13-inch, a difference of a factor of 4. For the heat treatment at 1165ºF, tests were conducted at 850ºF and 900ºF at hydrogen pressures of 3 ksi and 2 ksi. At the higher pressure, a factor of 4 was observed again. The lives of the 12-inch and the 13-inch materials were 3,697 hours and 15,327 hours, respectively. At the lower pressure and 900ºF, the respective lives were 7,490 hours and 31,450 hours, also a factor of 4 (see 172HFigure 51). The life ratios of the different materials in hydrogen were then quite similar to those observed in air. Superior creep strength was retained under hydrogen. This suggests a direct correlation in the creep range between resistance to HTHA and creep strength


35

12.0 EXPOSURE TIME OR STRESS

The results reported above quantify the effects of important variables on the hydrogen attack susceptibility of 2-1/4Cr–1Mo steel. While the trends are not surprising, the data developed were used to correlate behavior for the purposes of assessing the interaction of hydrogen pressure and stress, as described later. It was found that longer term (lower stress) tests tended to show increasing amounts of degradation at given hydrogen pressures, as shown in 173HFigure 52. However, it was recognized that the trend was complex and not easily quantified at the time. It was concluded from the MPC study that optimizing resistance to hydrogen attack requires PWHT at relatively high temperatures (1275oF). Hydrogen attack susceptibility appeared to be greatest in the Heat Affected Zone and varied from heat to heat due to inherent differences in creep resistance (consistent with cavity growth models). The data indicated that, for operating conditions of 825ºF and above, hydrogen attack might be anticipated in highly stressed components of 2-1/4Cr-1Mo given minimal PWHT. When higher operating temperatures are anticipated, a vanadium modified alloy is appropriate. Creep strength and resistance to hydrogen attack for vanadium modified heats developed and studied under MPC (64) were far superior to conventional 3-Chrome grades which do not contain vanadium, but which have been used in the past at higher temperatures than 2-1/4 Cr-1 Mo steel, but only at low stresses. API has assigned the 2-1/4Cr–1Mo-V alloy, developed under MPC‟s project, the same position on the Nelson Curve as the 3-Chrome grades. MPC exposed the base metal of its vanadium bearing grade for about 100,000 hours under high stress in 3000 psi hydrogen at 950ºF without failure. The carbon activity estimate for base metal of this grade is .01 and is generally consistent with the estimates made using methods reported in Appendix G by European investigators. For the HAZ of a weldment, higher values of the carbon activity would be appropriate.


36

13.0 EUROPEAN STUDIES

Recent European work greatly advanced modeling of attack-related phenomena. Abstracts and commentaries are attached in Appendix G. Researchers at Delft University (12-16) explored non-uniform cavitation processes on the grain size scale in polycrystalline aggregates. Analyses were made of the influence of the surrounding grains on development of grain boundary attack. Numerical results were explained in terms of two simplified models for the grain deformation–grain boundary cavitation interaction process. Building on their earlier work, the Delft investigators extended a cavity growth model in which diffusion and creep were coupled. They performed detailed numerical analyses of the possible stress states that may be encountered. The combined effect of internal methane and hydrogen pressures and applied stresses were computed. The technique was applied to predict failure in 2-1/4Cr–1Mo steels based on estimates of the free energies of the carbides present. The Delft work first indicated how attack develops in time. It was assumed that the cavities were distributed uniformly along the grain boundary facets. For the later work they noted carbides are not uniformly distributed over a grain facet and, more importantly, the carbides can have different compositions leading to significant differences in the methane pressure that may form. They noted too that, grain boundary cavitation would be slowed by surrounding grains, which deform by creep, only very slowly. Thus, the results of computer simulations of cavitation in a polycrystalline aggregates differed from those obtained for simple single cavity models. A two–dimensional polycrystal model of Van der Giessen and Tvergaard (15) was used to study the influence of carbide variations under different conditions. They showed that the growth rate of cavities in a creeping matrix is very strongly dependent on stress triaxiality, They concluded that void growth models may predict growth rates that are several orders of magnitude in error, depending on stress triaxiality and porosity. Application of the Delft model to 2-1/4Cr–1Mo (13) steel indicated that the time to failure is very sensitive to the type and composition of the carbides in the material since these determine the equilibrium methane pressure. The investigators at Delfy predicted the same for other materials. They also concluded that there would be strong sensitivity to heat treatment, consistent with the observations reported above for MPC‟s experiments. The Delft researchers‟ model confirmed the importance of stress. Stresses could be either residual or due to the applied loading. The effect of stress was related primarily to the methane pressure through a triaxiality parameter and a pseudo length parameter. In short, the Delft researchers concluded that there are two cavitation modes that can be covered by continuum damage relations. In one case, the grains can deform significantly, the cavitation concentrates primarily on facets perpendicular to the macroscopic maximum principal stress. In the other, creep deformation of the grain is negligible and cavitation develops uniformly on all facets. Recently, Schlogl and van der Geissen synthesized their results of modeling and finite element analysis into a coherent picture of the progress of hydrogen attack in a welded 2-1/4Cr–1Mo steel pressure vessel (77). Their story is valuable because, through attention to details, it addresses all the issues that may arise in any steel and component.


37

They note the presence of multiple carbides - in this case M7C3, M6C, and M2C. Since in low alloy steels, the carbides are not pure (i.e., binary, a metal and carbon), the thermodynamic driving force for methane formation is not known with precision. There is then significant uncertainty in the methane pressure calculation. Their approach is traditional; they write the equilibrium equation for the chemical reactions describing the breakdown of the carbides by hydrogen to form methane, e.g.,

[37]

For alloy steel with lower percentages of carbide forming elements than 2-1/4Cr–1Mo, the potential existence of M3C, essentially Fe3C, is of paramount concern, and one need not usually proceed much further if its presence is expected. The reason is that where cementite type carbides are found the free energy driving force for methane formation, and likely the thermodynamics of behavior will be dominated by that phase. While M3C type carbides may include molybdenum, chromium, manganese, or other elements, carbides of the M23C6, M2C, or M6C types are far more stable and are not expected to influence behavior. The process used by Schlogl, et a (77)l., is to estimate the Gibbs free energy by writing an equation for the equilibrium state of the above reaction as follows:

[38]

For the carbide M7C3 which they anticipate will be the “main” carbide of importance (one should always seek the least stable carbide present on the assumption that it will be most easily reacted and generate the highest methane pressure), the following form is used:

[39]

where: y1 and y2 are the respective fractions of chromium and iron in the carbide. One may rigorously proceed to develop the free energy of the ferrite:

[40]

and the potentials through partial derivatives:


38

[41]

and

[42]

However, Schlogl, et al., note that this is a blind alley. One must just estimate the free energies based on measurements (or guesses) of the carbide compositions and the equation for potentials above. Inserting the Gibbs free energies from all these estimates and the local hydrogen concentration (equivalent pressure) yields an estimate of the highest methane potential, its fugacity. Because methane is nonideal and fugacity is very high , especially at the lower temperatures of interest, one uses the relation noted earlier (Equation 14) to calculate the methane pressure of interest.

[43]

The study includes the usual concepts of damage due to coupled diffusional and creep void growth, i.e.

[44]

and the creep rate

[45]

A stress and temperature dependent “length” parameter is established, as noted earlier in Equation 24.

[46]

This damage model was applied to weld metal, coarse-grain HAZ, fine-grain HAZ and base metal. Each zone had different strengths, grain size, carbides, carbon activity and potential methane pressure. Propagation of damage was studied through the hydrogen content gradient through the wall (although an error was apparently made here in taking an essentially linear effective pressure gradient through the wall rather than using the linear gradient of hydrogen concentration). Methane pressure in equilibrium with the M7C3 of varying chromium content was estimated.


39

Computations of damage gradients through the wall were examined at various life fractions as a function of the three principal stress directions. In their example, damage accumulated most rapidly in HAZ‟s where creep strength was the lowest. Activities in Europe were spurred by an interest in predicting the behavior of advanced alloys under the most severe operating conditions envisioned for low alloy steels. The diffusion/creep interaction modeling carried the concepts of early void growth modelers such as Johnson, Shewmon and Odette to new levels of refinement. They provide a complete, integrated approach to the effects of carbide composition, applied stress, temperature, and nonideal gas behavior on time to failure. Limitations of the models are materials properties estimates, confirmation of estimates of carbon activity and the absence of a database that could support the conclusions. It was concluded that carbon activity in vanadium modified low alloy steels is so low that potential methane pressures would be easily tolerated.


40

14.0 NEW ANALYSIS OF MPC WORK

The results of MPC‟s extensive studies of 2-1/4Cr–1Mo steel were examined using the concept of methane pressure as the driving force to develop isochronous failure contours. For this work, a macrophenomenological model was used based on the following concepts:

1) The activity of carbon was estimated for the conditions of PWHT evaluated by MPC. The

values chosen were 0.03 and 0.06 for the heavy (1275ºF) and light (1165ºF) tempering of weldments , respectively.

2) The line for 2-1/4 Cr – 1 Mo steel currently in API RP 941 was assumed to be close to correct for at least 100,000 hours, unstressed service for the heavy tempering heat treatment. The corresponding methane pressures can be set equal to the design allowable stress for that temperature.

3) Methane pressure and applied stress were assumed to interact in producing grain boundary damage.

4) It was assumed that hydrogen alone could lead to failure through methane pressure development. The exponent for methane pressure dependence of unstressed failure time was estimated to be between about 4, an intermediate value between theories and rupture tests in air, absent a database failure behavior for of unstressed exposures in hydrogen.

5) Equations for stress-rupture lives observed in air were used. Typically, the exponent for stress-rupture behavior in air is between 6 and 10.

6) The relation that captures an interaction between damage in air and in hydrogen under stress was expressed as follows:

[47]

where: m = an exponent obtained by fitting the interaction of creep and methane pressure damage modes, tH,S = failure time under stress and hydrogen (methane), and tH, tS = failure times in hydrogen (methane) or under stress in air, individually.

7) For any heat treatment of the steel, carbon activity was assigned. For the hydrogen

pressure the corresponding methane pressure can be calculated. 8) Using the interaction exponent as a dummy variable, isochronous life predictions can be

compared to available test data from MPC , European or other test programs.

For a selected time to failure, a range of applied stresses was examined. For each stress, the failure time in air was calculated. Since the equations are all in closed form, the corresponding methane and hydrogen pressures to give the target life could be calculated . The results applying the approach above are shown for some long term MPC results (64) and the short term, high temperature tests of the European researchers (78) are plotted in 174HFigure 53 to 175HFigure 56. Applied stress is found to reduce the life in hydrogen in a complex fashion and to a degree dependent on creep strength in air. This is consistent with experience.


41

The best fit interaction exponents for the data, m, were found by trial, usually about in the range of 1/3 to 1/2. The approach was also applied successfully to very old results from work by Vitovec on C - 0.5% Mo steel. See 176HFigure 57. (24) The result of this exercise (curves in 177HFigure 53 to 178HFigure 57)suggests this is a reasonable approach to modeling performance given properties in air and some benchmarks for hydrogen performance. The pressure-applied stress contours trace the complex response to stress observed in MPC‟s program as described above. These pressure/stress contours suggest a starting point for developing life assessment tools and, at least in preliminary fashion, evaluating the effects of operation under variable conditions (over-heating excursions) as discussed in Appendix C and Appendix H.


42

15.0 DEVELOPMENT OF P/T LIMITS NEAR THE CREEP RANGE AND ABOVE

From the above, in the creep range at moderate applied stresses, long term hydrogen P/T limits may be estimated by finding for any temperature the hydrogen pressure corresponding to methane pressures on the order of the design allowable stress – i.e., a stress giving negligible creep. This would be consistent with void growth models. (Below the creep range, the limit may be set by the reaction rate estimating hydrogen solubility and carbon concentration in the matrix.) Examples are illustrated in the following: 179HFigure 58 focuses on the pressure relations at temperatures of interest in the creep range for carbon steel. The hydrogen pressures corresponding to the allowable stress limits at each temperature are cross plotted. The relatively constant pressure behavior described earlier and found on Nelson‟s curves may be seen to lie in the neighborhood of 100 psi. 180HFigure 59 shows the benefits of increased strength and reduced carbon activity that might be expected for a C – 0.5% Mo or similar steel. The limiting hydrogen pressure increases because of the increased strength and lower activity expected. Pressure limits for 1-1/4Cr–1/2Mo steel at carbon activity of 0.1 are found in 181HFigure 60. Pressure increases with decreasing temperature below the creep range. Two curves are shown for 2-1/4Cr–1Mo steel in 182HFigure 61 and 183HFigure 62. The first is for the heavily tempered (1275ºF, multiple cycles) variety. At typical operating temperatures, hydrogen pressures required for a clad, thick wall reactor seem consistent with the calculations. Increasing temperatures do not significantly reduce the calculated acceptable value. Material tempered at lower temperatures (1165ºF) would be expected to have higher carbon activity and more free carbon in solution ( 184HFigure 62) and thus reduced acceptable hydrogen partial pressures. However, the acceptable hydrogen partial pressures are only slightly reduced because these steels have enhanced creep strength. On the other hand, as a practical matter, the methane forming reaction rate at lower temperatures might be more rapid due to the increased amount of carbon remaining in solution. Vanadium bearing steels (185HFigure 63) would appear to have a significant hydrogen pressure margin for current designs based on an alloy with an estimated carbon activity of 0 .01.


43

16.0 A TECHNICAL BASIS

Inspection of the plots of methane pressure and temperatures for steels discussed above suggests the reasonableness of Nelson‟s placements. Depending on the activities selected, the pressure limits may vary somewhat, but are within narrow ranges consistent with experience. Relating position to strength was an idea of early investigators. However, they could not quantify the nonideal behavior of methane pressure, nor did they adequately consider materials properties including carbon activities and availability in the matrix. The plots below incorporate the concepts described above including the estimates for the effects of temperature and pressure on reaction rates. 186HFigure 64 to 187HFigure 67 provide “predictions” of where the P/T should be for carbon steel, carbon 0.5% Mo, 1–1/4 Cr alloy and 2–1/4 Cr alloy for reasonable combinations of the respective activities and creep strengths. As noted earlier, the creep strength basis at high temperatures breaks down where a different approach is needed where reaction rates are slow. Appendix C provides guidance about strengths that may be used to develop the limits or design allowable stresses may be used. The carbon activities sufficient to explain the behavior historically observed are as follows:

Table 3 – Carbon Activity Table

Steel Carbon Activity Assumption

Carbon 1 Fe3C C – 0.5% Mo .5 to .8 M3C

1Cr- or 1-1/4Cr - 1/2Mo .05 to .15 dissolved C 2-1/4Cr–1Mo .03 to .10 dissolved C

3Cr–1Mo .03 to .10 dissolved C V modified or high Cr .005 to .015 dissolved C

It has been assumed that carbon activity is related to carbon content through a multiplier in the range of unity justified in 188HFigure 37. Carbon activity depends on composition and heat treatment. This is true of strength as well. Thus, individual heats of the alloys will exhibit a range of performance, depending on composition and heat treatment. The notion being advanced here is that if the methane pressure generated is insufficient to cause an appreciable creep rate or deformation at the microscopic level, attack will not occur. This is generally consistent with all models for void growth and avoids concern for the details of diffusion processes (Appendix F and Appendix G.) Analytical models over the years incorporated creep rate in the calculation, but models have suffered from poor characterization of materials‟ strength parameters. They also tended to predict attack at too low temperatures, perhaps because nucleation and reaction rates are not adequately characterized with regard to carbon concentration and microstructure.


44

A simple solution to the problem of overestimating the tendency for hydrogen attack at low temperatures offered here is to examine the product of hydrogen and carbon contents. This concept was tested by assuming that Nelson‟s carbon steel line is roughly in the correct location. Hydrogen content on the attack line can be calculated and the critical product (see Equation 48) of hydrogen and carbon estimated. The fact that it is nearly constant along the “safe lines” reinforces the argument. The inferred reaction rates were applied to the other alloys. Low alloy steels will have higher temperature plateaus - i.e. higher hydrogen pressures at higher temperatures because the amount of dissolved carbon available to participate in the reactions is lower, see 189HFigure C-1. Nucleation rates are not so easily estimated and must be viewed as dependent on microstructure (coarse carbides and inclusions being detrimental).

In essence, at low temperatures, the reaction rate is adequate to sustain attack when:

[48]

where: [C] = dissolved carbon, [H] = dissolved hydrogen, and m = an exponent between 1 and 4. This approach leaves open the issue of exactly where the reaction occurs - i.e. on the carbide surface or at any other interface.


45

17.0 DISCUSSION

The topic of HTHA has been researched for over seventy years. The seemingly simple curves developed by G.A. Nelson (Appendix A) attempted to capture industry experience through the 1950‟s and 1960‟s. With the notable exception of C – 0.5% Mo steel, “the Nelson Curves” depicting satisfactory and unsatisfactory pressure-temperature regions have survived as reasonable to this day. The pressure-temperature curves he prepared were based on Nelson‟s judgment after considering reports about laboratory tests and plant experience. His judgment seems to have been quite good because much of the raw “data” he obtained might not stand up to close scrutiny. The information he was given by plant operators about service pressure and temperature, or the absence of damage should not be accepted at face value. This statement is made here because, even currently, we find it difficult to obtain accurate operating data and material damage assessments. This barrier to gathering reliable data is unlikely to have gotten worse since Nelson‟s time. While the data gathered by Nelson from the field were certainly suspect, laboratory work from his time on has not yielded conclusive information helpful in today‟s climate of interest in fitness-for-service, risk and life assessment. One can debate whether Nelson‟s success has been the result of remarkable insight, his conservatism or just that his curves have not been effectively tested in service because of large operating safety margins used in refineries. Certainly the success of API RP 941 stems in part from the fact the curves have been conservatively applied in regard to design and operation. Recently, however, a growing list of problems with C – 0.5% Mo steel underscored the lack of understanding of HTHA and heightened concern about lengthening exposure time and the curves for other steels. Also, now it is desired to push equipment to the limits of API RP 941 for economic reasons and to perform risk-based inspection. Appendix A provides a detailed discussion of the background of API RP 941 and the data gathered. It is worth pondering why seemingly little progress in understanding HTHA has been made since Nelson‟s times. We are still far from being able to make quantitative predictions about the behavior of steels subject to HTHA. A literature review done for this study has led to a long list of obstacles to establishing more appropriate operating limits for steels. The list in Appendix E should provide someone generally familiar with HTHA with additional insight into the nature of HTHA and the current state of affairs. Also refer to Appendix A and the references cited in API RP 941 to obtain additional technical foundation. The list in Appendix E explains why there is currently inadequate data about the specific effects of microstructure, carbon activity and hydrogen content. The inconvenient aspects of studying reaction kinetics and calculating methane pressure leaves engineers poorly equipped to consider the actual driving force for initiation and propagation of attack. This report has attempted to overcome shortcomings of prior investigations by focusing on quantifying the role of methane pressure on causing HTHA damage. To validate a model for HTHA it would be necessary to fashion a comprehensive test program and then determine for one steel the functional relations of all key variables in hydrogen attack. Such variables are stress, grain size, tensile and creep strengths, carbon content, temperature, methane pressure, microstructure, etc. Only MPC‟s stress rupture results for 2–1/4Cr–1Mo in high-pressure hydrogen approach being such a database, but they fall short. The set of isochronous lines for carbon steel in API RP 941 are not a satisfactory starting point.


46

Finally, as noted in Appendix A, Nelson reportedly created his curves by applying a safety margin of 30oF to 50ºF to his curves for reported damage. Also, Nelson referred to depicting average material behavior. Exactly what the lines were meant to signify is unclear, but they seem to be appropriate, with one notable exception, for exposures absent high applied stresses. Today some suggest the correct C – 0.5% Mo curve should be only 50ºF above Nelson‟s carbon steel line. In other words, if Nelson applied a 50ºF safety margin for carbon steel, the newly suggested C – 0.5% Mo steel curve would sit essentially on Nelson‟s unfactored C steel curve. Despite their possible shortcomings, Nelson‟s lines are still a good guide to trends and are a useful basis for qualitative judgments.


47

18.0 FINAL COMMENT

The concept of a simple boundary between safe and unsafe operating conditions in hydrogen for common alloys, of the type depicted by the Nelson Curves should not be expected. Certainly material composition, heat treatment and stress are well accepted as variables that influence behavior. Operating above the current lines can lead to short life and time dependent failure.

Experience shows damage accumulation is time dependent. However, the methods of detection and quantification of damage are so inadequate, operating conditions so poorly recorded, failure analyses so cursory and materials characterization so primitive, that life prediction is on shaky grounds today. Studies that can now be defined utilizing knowledge of methane pressure to get at the kinetics are needed. These would permit developing correlations useful for design and for fitness-for-service. Hydrogen content, potential methane pressure, and relative reaction rate based on standard kinetics will be needed to accurately predict behavior. A procedure for evaluation of recommended equipment would be as follows:

1) Using appropriate diffusivities, solubilities and fluxes, calculate local through wall effective

hydrogen pressures. 2) Using methane pressure vs. hydrogen pressure reference lines for each alloy/heat

treatment and the corresponding appropriate activity of carbon, estimate the potential methane pressure at the operating temperature.

3) Using the approach presented here, determine if conditions are in a range that may be potentially damaging.

4) Using the methods suggested herein, estimate the applied stress interaction and the damage rate.

5) See Appendix C where operating conditions may be in the reaction rate limited regime.

Note: It is not known if a linear life fraction rule can be justified for this application. However, for small excursions, if conservatively applied, the technique could be used. The insights gained in preparing this Technical Basis Document suggest the outlines of studies that might profitably be conducted to test the concepts advanced herein and provide important information for Fitness for Service evaluation, inspection planning and risk based equipment management. To begin with, the industry would benefit by developing means of quantifying the progress of hydrogen attack damage. This would facilitate communication, data gathering, documentation, reporting information to API, damage modeling and eventually predictive tools. This goal would be more readily accomplished in the context of a test program that included interrupted testing of specimens exposed to hydrogen on only one side. There is little reliable information on the time dependence of damage. Monitoring damage after one side exposures would provide confidence that the approaches suggested herein are on the right track as well as information valuable for setting inspection intervals.

Now that there is a focus on role of internal methane pressure on shortening stress-rupture life, more effective test plans can be developed. Great benefit could be derived from planning tests wherein methane pressure, rather than hydrogen pressure, is the experimental variable. It is essential that this be done for each alloy of interest.. The complex interaction of applied stress and methane pressure may differ among the alloys of interest.


48

All test programs should include, in addition to the usual microstructural characterization of the materials, a convenient measure of the availability or reactivity of carbon. The method might be something as simple as a separate test to measure decarburization rate or determining the amount of carbon left in solid solution after tempering/PWHT. Finally, monitoring of hydrogen diffusion through operating equipment is long overdue. There are many things we have yet to learn. The few items listed here would provide a wealth of necessary information. The following are some thoughts about a test program and actions to be taken. We now recognize that details of attack are sensitive to many variables. Conclusions about behavior of materials susceptibility should be based on tests at pressures and temperatures as close as practical to operating conditions. The insights gained in preparing this Technical Basis Document suggest the outlines of studies that might profitably be conducted to test the concepts advanced herein and provide important information for Fitness-for-Service evaluation, inspection planning and risk based equipment management.


49

19.0 OVERVIEW AND CONCLUSION

While there are many shortcomings with regard to the historical records of operating conditions that have led to attack the general trends and positions of the operating limit lines provided by API RP 941, have not been seriously called into doubt , aside from those provided for the 0.5% Mo steels,.

The kinetics of the methane reaction suggest an explanation for the shapes of the curves in API RP 941. At temperatures above the “knee”, reactions may be rapid, equilibrium methane pressures can be reached if voids and cracks initiate. At temperatures below the knee, a pressure “plateau” is observed because rates of nucleation and methane formation are low.

The rate of methane formation in steel is expected to increase with dissolved hydrogen content, carbon content, temperature and the available nucleation sites for bubbles or fissures.

When temperature, hydrogen content and carbon activity in the steel produce methane at pressures in excess of the material‟s tensile or creep strength, attack will proceed over a period of time. This concept was proposed years ago by many investigators.

Measurement of carbon activity in steel does not appear to be practical, but some reasonable estimates of the activity based on the literature, past performance and free carbon concentration are provided herein to enable estimates of methane pressure.

The methane pressures that may develop in cavities or cracks at the safe operating limits in API RP 941 appear to be consistent with material strength properties.

An approach has been developed whereby stress-rupture tests in hydrogen can be used as benchmarks to estimate isochronous failure curves for the hydrogen pressure-applied stress interaction.

Sophisticated void growth models (Appendix G) are useful in explaining trends generally conforming to the curves in API RP 941. However, implementing the models requires numerous material properties that are not available and many assumptions that make quantitative implementation impractical at this time.

Simplified equations for reaction rate based parameters of the and type (Appendix F) depict the trend of behavior of limited data for C - 0.5% Mo steel, upon which they were based. However, they are difficult to justify and implement for other low alloy steels.

Austenitic and ferritic stainless steel weld overlay or cladding can prevent or retard hydrogen attack by lowering hydrogen content in the backing steel. High-chrome ferritic steels are less effective than austenitic, but still potentially very helpful. Benefits of overlay or cladding decrease with increasing backing steel thickness (Appendix D).


50

Progress in understanding hydrogen attack has been hampered by problems that are not easily overcome. These include the difficulties of measuring carbon activity, quantifying the degree of attack, characterizing microstructure (nucleation sites) and designing conditions for useful experiments. Available test results are not adequate to enable quantitative prediction of behavior. Recent improvement in understanding the phenomenon should allow development of a test program to establish safe operating limits, a Fitness-For-Service methodology and a better understanding of issues needed for Risk-Based Inspection.


51

20.0 REFERENCES 1 API Publication 945, A Study of the Effects of High Temperature, High Pressure

Hydrogen on Low-Alloy Steels, American Petroleum Institute, Washington, D.C., 1975 (out of print).

2 API Publication 940, Steel Deterioration in Hydrogen, American Petroleum Institute, Washington, D.C., 1967 (out of print).

3 API Refinery Corrosion Committee Survey, 1957. 4 Steels for Hydrogen Services at Elevated Temperatures and Pressure in Petroleum

Refineries and Petrochemical Plants, American Petroleum Institute, Publication 941, 3rd Ed., 1983.

5 G. A. Nelson, “Hydrogenation Plant Steels,” Proceedings, 1949, Volume 29M, American Petroleum Institute, Washington, D.C., pp. 163-174.

6 G. A. Nelson, “Operating Limits and Incubation Times for Steels in Hydrogen Service,” Proceedings, 1965, Volume 45, American Petroleum Institute, Washington, D.C. pp. 190-195.

7 F. K. Naumann, “Influence of Alloy Additions to Steel upon Resistance to Hydrogen Under High Pressure,” Technische Mitteilungen Krupp, Volume I, Number 12, 1938, pp. 223-234.

8 B. L. Chao, G. R. Odette, and G. E. Lucas, “Kinetics and Mechanisms of Hydrogen Attack in 2.25Cr-1Mo Steel,” ORNL/Sub/82-22276/01, August 1988.

9 T. Sakai, T. Nomura, and E.H. Niccolls, “The Basis and Application of the “PW “ Parameter for High Temperature Hydrogen Attack”.

10 H.M. Shih and H.H. Johnson, A Model Calculation of the Nelson Curves for Hydrogen Attack, Acta Metall,. 30, p. 537, 1982.

11 S.S. Vagarali and G.R. Odette, “Analysis of Hydrogen Attack on Pressure Vessel Steels,” Fossil Energy Materials Program Quarterly Progress Report for Period Ending March 31, 1981.

12 M.W.D. van der Burg, E. van der Giessen, and V. Tvergaard, A Continuum Damage Analysis of Hydrogen Attack in a 2.25 Cr – 1 Mo Pressure Vessel, Material Science and Engineering.

13 M.W.D. van der Burg, E. van der Giessen, and R.C. Brouwer, Investigation of Hydrogen Attack in 2.25 Cr – 1 Mo with a High Triaxiality Void Growth Model, Acta Mettal. 44, p. 505, 1996.

14 M.W.D. van der Burg and E. van der Giessen, “Non-Uniform Hydrogen Attack and the Role of Interaction With Creep,” Material Science and Engineering, A220, p. 200, 1996.

15 E. van der Giessen and V. Tvergaard, Acta Metall, 42 (1994) 959. 16 E. van der Giessen, M.W.D. van der Burg, A. Needleman and V. Tvergaard, Journal.

Mech. Phys. Solids 43, 123 (1995). 17 R. E. Allen, R. J. Jansen, P. C. Rosenthal, and F. H. Vitovec, “The Rate of Irreversible

Hydrogen Attack of Steel at Elevated Temperatures,” Proceedings, 1961, Volume 41, American Petroleum Institute, New York, pp. 74-84.

18 H.J. Grabke and E. Martin, Arch. Eisenheuttenwes, Volume 44, p. 837, 1972. 19 G.R. Odette and S.S. Vagrali, Metallurgical and Materials Transactions, Volume 12A, p.

2071, 1981.


52

20 G.R. Odette and S.S. Vagrali, „An Equation-Of-State for Methane for Modeling Hydrogen Attack In Ferritic Steels‟, Metallurgical and Materials Transactions, Volume 13A, p. 299, 1982.

21 R.E. Allen, R.J. Jansen, P.C. Rosenthal, and F.H. Vitovec, “Analysis of Probable Mechanisms of High-Temperature Hydrogen Attack of Steel,” Presentation of American Petroleum Institute‟s Division of Refining, May 15, 1962.

22 G.H. Geiger, O.F. Angeles: Study of Effects of High-Temperature, High-Pressure Hydrogen on Low-Alloy Steels, American Petroleum Institute, Publication 945, 1975.

23 W. Geller and T. Sun, “Influence of Alloy Additions on Hydrogen Diffusion in Iron and Contribution to the System lron-Hydrogen,” Arch. Eisenhuttenw 21, pp 423-430, 1950.

24 F. H. Vitovec, “The Growth Rate of Fissures During Hydrogen Attack of Steels” Proceedings, 1964, Volume 44, American Petroleum Institute, Washington, D.C., pp. 179-188.

25 M. McKimpson and P.G. Shewmon, Metallurgical and Materials Transactions, 1981, Volume 12A, p. 825.

26 T. Boellinghaus, H. Hoffmeister, and C. Middel, "Scatterbands for Hydrogen Diffusion Coefficients in Low and High Alloyed Steels with An Austenite Decomposition Microstructure and High Alloyed Steels with an Austenitic Microstructure at Room Temperature", IIW Doc. IX-1812-95

27 T. Boellinghaus, H. Hoffmeister, and A Dangeleit, " A Scatterband for Hydrogen Diffusion Coefficients in Microalloyed and Low Carbon Structural Steels", IIW Doc. IX-1767-94

28 T. P. Groeneveld and A. R. Elsea, “Permeation of Hydrogen at Elevated Temperatures and Pressures Through Stainless Steel Overlayed 2-1/4Cr-IMo Steel”, Proceedings of American Petroleum Institute, Division of Refining, Volume (Midyear Meeting, May 9-12, 1977, Chicago, IL, pp. 7-16. API, Washington. D.C., 1977.

29 “Report on the Effect of Stainless Steel Weld Overlay or Cladding on Hydrogen Attack of Underlying Steel,” Materials Properties Council, New York, September 1984.

30 M. H. Armbruster, “The Solubility of Hydrogen at Low Pressure in Iron, Nickel, and Certain Steels at 400 to 600C” J. Am. Chem. Soc. 65, p 1043, 1943.

31 ASME Boiler and Pressure Vessel Code Section II Part D 2004 32 J.P. Hirth and H.H. Johnson, CORROSION, Volume 32, Number 1, p.3, January 1976. 33 P.G. Shewmon and Z.S. Yu, Advanced Material for Pressure Vessel; Vessel Service

with Hydrogen at High Temperature and Pressures, ASME, New York, 1982, pp. 85-92. 34 P.G. Shewmon, Material Science and Technology, Volume 1, 1985, pp. 1-8. 35 P.G. Shewmon, Acta Metall, Volume 35, 1987, pp. 1317-24. 36 P.G. Shewmon, Metallurgical and Materials Transactions, Volume 7A, p. 279, 1976. 37 R. Pishko, M. McKimpson, and P.G. Shewmon, Metallurgical and Materials

Transactions, 1979, Volume 10A, p. 887. 38 B. Panda and P.G. Shewmon, Metallurgical and Materials Transactions, 1984, Volume

15A, p. 487. 39 T.A. Parthasarathy, P.G. Shewmon: Metallurgical and Materials Transactions, Volume

15A, 1984, pp. 2021-27. 40 T.A. Parthasarathy, Mechanisms of Hydrogen Attack of Carbon and 2.25 Cr – 1 Mo

Steels, Acta Metall. 9, p. 1673, 1985. 41 T.A. Parthasarathy, H.F. Lopez, and P.G. Shewmon, Metallurgical and Materials

Transactions, Volume 16A, 1985, pp. 1143-49.


53

42 T.A. Parthasarathy, P.G. Shewmon: Metallurgical and Materials Transactions, Volume 18A, 1987, pp. 1309-12.

43 P.G. Shewmon, Y.H. Xue: Metallurgical and Materials Transactions, Volume 22A, 1991, pp. 2703-07.

44 G. Sundarajan and P.G. Shewmon, Metallurgical and Materials Transactions, 1980, Volume 11A, p. 509.

45 G. Sundarajan and P.G. Shewmon, The Kinetics of Hydrogen Attack of Steels, Metallurgical and Materials Transactions 12A, p. 1761, 1981.

46 R. Raj, M.S. Shih, and H.H. Shewmon, Scripta Metal, Volume 11, p. 839, 1977. 47 T.J. Hakkarainen, J. Wanagel, and J. Li, C-Y., Metallurgical and Materials Transactions,

1980, 11A, 2035. 48 Z.J. Su, D. Stone, J. Wanagel, and C.Y. Li: Res Mechanical, Volume 13, 1985, pp. 243-

50. 49 J. Wanagel, T.J. Hakkarainen, and C.Y. Li: Application of 2 ¼ Cr – 1 Mo Steel for Thick

Walled Pressure Vessels, ASTM STP-755, Eds. G.S. Sangdahl and M. Semchyshen, ASTM, 1982, pp. 93-108.

50 Temper Embrittlement and Hydrogen Embrittlement in Pressure Vessel Steels, Japanese Pressure Vessel Research. Committee., Report #2, May 1979.

51 JPVRC Report No 2: Iron and Steel Institute of Japan, May, 1979. 52 M. Hasegawa and S. Fujinaga, “Attack of Hydrogen on Oil Refinery Steels” Tetsu To

Hagane, 1960, Volume 46, Number 10, pp. 1349-1352. 53 R. Chiba, K. Ohnishi, K. Ishii, and K. Maeda, “Effect of Heat Treatment on the

Resistance of C-0.5Mo Steel Base Metal and Its Welds to Hydrogen Attack,” 1985 Proceedings, Refining Department, Volume 64, American Petroleum Institute, Washington, D.C. pp. 57-74.

54 T. Ishiguro, K. Kimura, T. Hatakeyama, T. Tahara and K. Kawano, “Effect of Metallurgical Factors on Hydrogen Attack Resistance in C-0.5Mo,” presented at the Second International Conference on Interaction with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, The Materials Properties Council, Inc., Vienna, Austria, October 19-21,1994.

55 H. Ishizuka and R. Chiba, Tetsu-to-Hagane, Volume 56, p. 93, 1970. 56 K. Ishii, K. Maeda, R. Chiba, and K. Ohnishi, “Intergranular Cracking of C-0.5Mo Steel in

a Hydrogen Environment at Elevated Temperatures and Pressures,” 1984 Proceedings, Refining Department, Volume 63, American Petroleum Institute, Washington, D.C. pp. 55-64.

57 Kawano, “Effect of Metallurgical Factors on Hydrogen Attack Resistance in C-0.5Mo Steel”, presented at the Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, The Materials Properties Council, Inc., Vienna, Austria, October 19-21, 1994.

58 G. R. Prescott, “History and Basis of Prediction of Hydrogen Attack of C-0.5Mo Steel,” a state-of-the-art review presented at Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, The Materials Properties Council, Inc., Vienna, Austria, October 19-21,1994.

59 T. Sakai and H. Kaji, Tetsu to Hagane, 1980, 66, 1133. 60 M. Hasegawa and S. Nomura, Journal - Iron Steel Institute of Japan, 1977, Volume 7, p.

187.


54

61 A. R. Ciuffreda, N. B. Heckler, and E.B. Norris, “Stress Rupture Behavior of Cr-Mo Steels in a High Pressure Hydrogen Environment,” (ASME H00227/MPC-18), American Society of Mechanical Engineers, New York, 1982.

62 A.R. Ciuffreda, N.B. Heckler, and E.B. Norris, “Advanced Material For Pressure Vessel Service with Hydrogen at High Temperatures and Pressures,” ASME, N.Y., MPC-18, 1982, pp. 53-68.

63 E. B. Norris and E. A. Sticha, “Effect of Hydrogen on the Stress-Rupture Strength of 2-1/4Cr-1Mo Steel,” Metal Properties for the Petroleum and Chemical Industries (ASME G00103/MPC-2), American Society of Mechanical Engineers, New York, 1976, pp. 590-592.

64 C. Lundin, K. Khan, P. Liu, and M. Prager, “Study of Hydrogen Attack Susceptibility of 2 ¼ Cr - 1Mo Steel”, Proceedings Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, October 19-21, 1994.

65 Yukio Tomita, et al., Properties Evaluation of 0.5 Mo Steel Damaged by Hydrogen Attack, TG6 Activity of Subcommittee on Hydrogen Embrittlement, Materials Division, The Japan Pressure Vessel Research Council.

66 R.J. Steuber and G.H. Geiger “Hydrogen Attack of Steel Under Dynamic Exposure Conditions”

67 E. A. Sticha, “Tubular Stress-Rupture Testing of Chromium-Molybdenum Steels with High-Pressure Hydrogen,” Journal of Basic Engineering, December 1969, Volume 91, American Society of Mechanical Engineers, New York, pp. 590-592.

68 Y. Tomita and H. Yamamoto, Properties Evaluation of 0.5Mo Steel Damaged by Hydrogen Attack, PVP – Volume 380, pp. 1–11, 1998.

69 H. Tsubakino, J. Kitasaka, and K. Yamakawa, Hydrogen Attack and Its Monitoring by an Electrochemical Method, PVP – Volume 380, pp.25-30, 1998.

70 T. Nomura, K. Murayama, and M. Matsushita, “Sample Test Results of All Liquid Phase Hydrogen Attack,” Japan Energy Corporation, presented to the API Task Group on Hydrogen Deterioration of Steels, May 1994.

71 Tohru Nomura and Takuichi Imanaka, Ultrasonic Detection of Hydrogen Attack in Low Alloy Steel, 1994 ASME Pressure Vessels and Piping Division Conference June 19-23, 1994, Minneapolis, Minnesota.

72 K. Kawano, K. Hattori, H. Yamamoto, F. Sakota, H. Okada, T. Tahara, H. Tanaka, and T. Ishiguro, “UT and Metallurgical Evaluations on Hydrogen Attacked C-1/2 Mo Steel Pressure Vessels,” PVP – Volume 239, pp 129-138, 1992.

73 K. Hattori and S. Aikawa, “Scheduling and Planning Inspection of C-0.5Mo Equipment Using The New Hydrogen Attack Tendency Chart,” PVP Volume 239/MPC-Volume 33, Serviceability of Petroleum Process and Power Equipment, ASME, 1992.

74 K. Hattori, “An Advanced Hydrogen Attack Tendency Chart as a Life Assessment Strategy for C-0.5Mo Equipment,” PVP – Volume 359, pp. 333-338,1997.

75 L.C. Chen, Stress Assisted Hydrogen Attack Cracking in 2.25 Cr - 1Mo Steels, Ph.D. Thesis, The Ohio State University, 1993.

76 L.C. Chen and P.G. Shewmon, Metallurgical and Materials Transactions, 1995, 26A, 2317.

77 S.M. Schlogl and E. van der Giessen hydrogen Attack in a Welded Reactor" submitted to Journal de Physique, 2004

78 P. Castello et al. "Creep of 2.25 Cr-1 Mo Tubular Test Pieces under Hydrogen Attack Conditions" IAM/JRC of EC Petten Netherlands


55


56

21.0 FIGURES

FIGURE 1 - VOIDS INDICATING HTHA INITIATION ON GRAIN BOUNDARIES ARE USUALLY ASSOCIATED WITH CARBIDES AT EARLY STAGES AT (10000X)

FIGURE 2 – MODELS OF VOID GROWTH INCLUDE INITIATION AT EQUALLY SPACED SPHERICAL OR LENTICULAR VOIDS THAT GROW DUE TO THE JOINT ACTION OF INTERNAL METHANE

PRESSURE ,P, AND APPLIIED STRESSES, S AND T.


57

FIGURE 3 – INTERGRANULAR FRACTURE SURFACES DISPLAY SMALL VOID SIZE AND DISTRIBUTION (10000X)


58

FIGURE 4 – SUBMICRON SIZE VOIDS COALESCE AND COVER THE INTERGRANULAR FRACTURE SURFACE


59

FIGURE 5 – VOIDS OFTEN INITIATE AT COARSE CARBIDES AND ON FERRITE/BAINITE BOUNDARIES


60

FIGURE 6 – EFFECT OF TEMPERATURE ON THE CALCULATED EQUILIBRIUM METHANE PRESSURE IN THE PRESENCE OF CEMENTITE AT UNIT CARBON ACTIVITY AND 1000 PSI

PRESSURE OF HYDROGEN


61

FIGURE 7 – EFFECT OF TEMPERATURE ON THE FORWARD METHANE FORMATION REACTION RATE AT 200 AND 2000 PSI HYDROGENPARTIAL PRESSURE


62

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

63

FIG

UR

E 8

– O

PER

ATI

NG

LIM

ITS

FOR

CO

MM

ON

STE

ELS

USE

D IN

PET

RO

LEU

M IN

DU

STR

Y PR

ESSU

RE

VESS

EL S

ERVI

CE

AS

PRES

ENTE

D IN

API

RP

941


64

FIGURE 9 – THE SAME METHANE FORMATION REACTION RATES OCCUR IN HIGH AND LOW PRESSURE REGIMES DEPENDING ON CARBON ACTIVITY AND TEMPERATURE CONSISTENT

WITH THE RELATIVELY HORIZONTAL OR FLAT PORTIONS SHOWN ON THE CURVES FOR EACH ALLOY IN API RP 941 (SEE FIGURE 8)


65

FIGURE 10A – NON IDEAL GAS FUGACITY-PRESSURE RELATION FOR METHANE


66

FIGURE 10B – EFFECT OF CARBON ACTIVITY ON THE CALCULATED EQUILIBRIUM METHANE PRESSURE IN THE PRESENCE OF CEMENTITE AT VARIOUS TEMPERATURES


67

FIGURE 11 – SOLUBILITY OF HYDROGEN IN FERRITE AS A FUNCTION OF TEMPERATURE AT INDICATED HYDROGEN PRESSURES (PSIA)


68

FIGURE 12 – TEMPERATURE-HYDROGEN PRESSURE COMBINATIONS FOR INDICATED HYDROGEN CONTENTS (PPM). LOW CONCENTRATIONS OF HYDROGEN MAY BE RESPONSIBLE

FOR HYDROGEN ATTACK AS SHOWN IN API RP 941

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

69

FIG

UR

E 13

– R

EPO

RTE

D D

IFFU

SIVI

TIES

FO

R H

YDR

OG

EN IN

IRO

NS

AN

D S

TEEL

S VA

RY

GR

EATL

Y

DEP

END

ING

ON

CO

MPO

SITI

ON

S, M

ICR

OST

RU

CTU

RE

AN

D T

EST

CO

ND

ITIO

NSA

S SH

OW

N IN

AN

IIW

SU

RVE

Y


70

FIGURE 14 – REPORTED DIFFUSIVITIES FOR HYDROGEN IN IRONS AND STEELS VARY GREATLY DEPENDING ON COMPOSITIONS, MICROSTRUCTURE AND TEST CONDITIONS FOR

IRONS AND STEELS AS SHOWN IN AN IIW SURVEY

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

71

FIG

UR

E 15

– R

EPO

RTE

D D

IFFU

SIVI

TIES

FO

R H

YDR

OG

EN IN

LO

W A

LLO

Y ST

EELS

VA

RY

GR

EATL

Y D

EPEN

DIN

G O

N C

OM

POSI

TIO

NS,

M

ICR

OST

RU

CTU

RE

AN

D T

EST

CO

ND

ITIO

NS

AS

SHO

WN

IN A

N II

W S

UR

VEY

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

72

FIG

UR

E 16

– R

EPO

RTE

D D

IFFU

SIVI

TIES

FO

R H

YDR

OG

EN IN

AU

STEN

ITIC

,HIG

H A

LLO

YSTE

ELS

VAR

Y G

REA

TLY

DEP

END

ING

ON

C

OM

POSI

TIO

NS,

MIC

RO

STR

UC

TUR

E A

ND

TES

T C

ON

DIT

ION

S A

S SH

OW

N IN

AN

IIW

SU

RVE

Y


73

FIGURE 17 – REPORTED DIFFUSIVITIES FOR HYDROGEN IN HIGH CHROME STEELS VARY GREATLY DEPENDING ON COMPOSITIONS, MICROSTRUCTURE AND TEST CONDITIONS AS

SHOWN IN AN IIW SURVEY


74

FIGURE 18 – CALCULATED RELATIVE HYDROGEN PRESSURE IMMEDIATELY BEHIND 3/16-IN. AUSTENITIC OR HIGH CHROMIUM (>12%) FERRITIC STAINLESS STEEL CLAD OR OVERLAY. PRESSURE REDUCTION DECREASES WITH INCREASING BACKING STEEL THICKNESS FOR

INTEGRAL CLAD OR OVERLAY ON STEEL


75

FIGURE 19 – CALCULATED REDUCED HYDROGEN PRESSURE IMMEDIATELY BEHIND HIGH CHROMIUM (>12%) FERRITIC STAINLESS STEEL CLAD OR OVERLAY. PRESSURE REDUCTION

DECREASES WITH INCREASING BACKING STEEL THICKNESS FOR INTEGRAL CLAD OR OVERLAY ON STEEL


76

FIGURE 20 – CALCULATED REDUCED HYDROGEN PRESSURE IMMEDIATELY BEHIND AUSTENITIC STAINLESS STEEL CLAD OR OVERLAY. PRESSURE REDUCTION DECREASES WITH INCREASING BACKING STEEL THICKNESSINTEGRAL CLAD OR OVERLAY ON STEEL


77

FIGURE 21 – COMPARISON OF RESULTS IN FIGURES 19 AND 20 FOR AUSTENITIC AND FERRITIC CLAD OR OVERLAY


78

FIGURE 22 – FREE ENERGY REDUCTION ASSOCIATED WITH METHANE FORMATION FROM CEMENTITE AND GRAPHITE STANDARD STATES. REDUCTION IS GREATER AT LOW

TEMPERATURES


79

FIGURE 23 – SLOPE OF INCREASE IN EQUILIBRIUM METHANE PRESSURE WITH INCREASING HYDROGEN PRESSURE DECREASES DUE TO NONIDEAL GAS BEHAVIOR OF METHANE.

PRESSURE INCREASES WITH DECREASING TEMPERATURE


80

FIGURE 24 – ISOCHRONOUS LINES FOR HYDROGEN ATTACK DAMAGE TO CARBON STEEL AS DEPICTED IN API RP 941 FOR 100, 200, 1000, 10000 HOUR AND INDEFINITE LIVES


81

FIGURE 25 – CALCULATED EQUILIBRIUM METHANE PRESSURES FOR THE ISOCHRONOUS LIVES SHOWN IN FIGURE 24 PLOTTED VERSUS RECIPROCAL OF TEMPERATURE


82

FIGURE 26 – RELATION BETWEEN EQUILIBRIUM METHANE PRESSURE AND HYDROGEN PARTIAL PRESSURE FOR INDICATED CARBON ACTIVITIES ASSOCIATED WITH A CEMENTITE

TYPE PHASE AT 450oF


83


TYPE PHASE AT 500oF


84


TYPE PHASE AT 550oF


85


TYPE PHASE AT 600OF


86


TYPE PHASE AT 650OF


87


TYPE PHASE AT 700OF


88


TYPE PHASE AT 750OF


89


TYPE PHASE AT 800OF


90


TYPE PHASE AT 850OF


91


TYPE PHASE AT 900OF


92


TYPE PHASE AT 1000OF


93

FIGURE 37 – CARBON ACTIVITY AT LOW CARBON CONTENTS IN LOW-ALLOY STEEL MAY BE ESTIMATED AS EQUAL TO CARBON CONTENT


94

FIGURE 38 – EFFECT OF CARBON ACTIVITY ON THE CALCULATED EQUILIBRIUM METHANE PRESSURE IN ALLOY STEELS WITH CARBON STANDARD STATE ASSUMED TO BE GRAPHITE


95

FIGURE 39 – RELATION BETWEEN EQUILIBRIUM METHANE PRESSURE AND HYDROGEN PARTIAL PRESSURE FOR INDICATED CARBON ACTIVITIES ASSOCIATED WITH CARBIDES

MORE STABLE THAN CEMENTITE FOUND IN LOW ALLOY STEELS AT 450OF


96




97




98




99




100




101




102


MORE STABLE THAN CEMENTITE AT 900OF


103

FIGURE 47 – CALCULATED RELATIVE RATES OF ATTACK THROUGH WALL THICKNESS BASED PV AND PW MODELS OR THE METHANE FORMATION REACTION RATE. EXAMINATION OF

MATERIALS EXPOSED ON ONE SIDE IN THE LABORATORY AND SERVICE SUGGESTS DEPTH OF PENETRATION DECREASES LESS RAPIDLY THAN PREDICTED BY PV OR PW MODELS


104

FIGURE 48 – EFFECT OF STRESS ON RUPTURE LIVES FOR A SINGLE HEAT OF WELDED 2 1/4 CR 1 MO STEEL TEMPERED AT 1165oF OR 1275oF AND TESTED UNDER 3000 PSI HYDROGEN AT

INDICATED TEMPERATURES IN MPC STUDY


105

FIGURE 49 – EFFECT OF TEMPERATURE ON STRESS RUPTURE LIVES FOR A SINGLE HEAT OF WELDED 2 1/4 CR 1 MO STEEL AND TESTED UNDER 3000 PSI HYDROGEN AT INDICATED

STRESSSES IN MPC STUDY


106

FIGURE 50 – EFFECT OF HYDROGEN PRESSURE ON STRESS RUPTURE LIVES OF 2 1/4 CR 1 MO STEEL. TESTS WERE ON WELDED SAMPLES OF TWO HEATS OF DIFFERENT THICKNESS


107

FIGURE 51 – COMPARISON OF STRESS RUPTURE LIVES IN HYDROGEN OF TWO HEATS OF WELDED 2 1/4 CR 1 MO STEEL (IDENTIFIED AS 1 AND 2) DISCLOSED THAT THE TENDENCY FOR

LONGER LIVES FOR ONE RELATIVE TO THE OTHER, THAT HAD BEEN OBSERVED IN AIR, PERSISTED IN HYDROGEN


108

FIGURE 52 – RELATIVE LIFE (AS COMPARED TO TESTS IN AIR) FOR MATERIALS TESTED IN HYDROGEN DECREASED WITH DECREASING APPLIED STRESS

(INCREASE LIFE EXPECTED IN AIR)


109

FIGURE 53 – PREDICTED ISOCHRONOUS RUPTURE CURVES FOR INDICATED HOURS SHOWING INTERACTION OF EQUILIBRIUM METHANE PRESSURE AND EXTERNALLY APPLIED STESSES

FOR WELDED 2 1/4 CR 1 MO STEEL


110

FIGURE 54 – PREDICTED ISOCHRONOUS RUPTURE CURVES FOR INDICATED HOURS SHOWING INTERACTION OF HYDROGEN PARTIAL PRESSURES CORRESPONDING TO METHANE

PRESSURES IN FIGURE 53 AND EXTERNALLY APPLIED STESSES FOR WELDED 2 1/4 CR 1 MO STEEL


111

FIGURE 55 – PREDICTED ISOCHRONOUS CURVES IN HOURS SHOWING INTERACTION OF EQUILIBRIUM METHANE PRESSURE AND APPLIED STESSES IN EUROPEAN TUBE TESTS AT

1112F ON 2 1/4 CR 1 MO STEEL


112

FIGURE 56 – PREDICTED ISOCHRONOUS CURVES IN HOURS SHOWING INTERACTION OF HYDROGEN PRESSURE CORRESPONDING TO METHANE PRESSURES IN FIGURE 55 AND

APPLIED STRESSES IN EUROPEAN TUBE AT 1112F TESTS ON 2 1/4 CR 1 MO STEEL


113

FIGURE 57 – PREDICTED ISOCHRONOUS CURVES FOR INDICATED HOURS SHOWING INTERACTION OF HYDROGEN PRESSURE AND EXTERNALLY APPLIED STESSES IN TESTS AT

1000F ON CARBON-0.5% MO STEEL


114

FIGURE 58 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR CARBON STEEL CONTAINING CEMENTITE

AT UNIT CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT EACH TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE

CREEP STRENGTH GOVERNED DESIGN ALLOWABLE STRESS AND ON THE VERTICAL AXIS FOR THAT POINT AN ESTIMATE OF THE SAFE WORKING HYDROGEN PRESSURE


115

FIGURE 59 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR CARBON-0.5% MO STEEL CONTAINING CEMENTITE AT 0.5 CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT EACH

TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE CREEP STRENGTH GOVERNED DESIGN ALLOWABLE STRESS AND ON THE VERTICAL AXIS

FOR THAT POINT AN ESTIMATE OF THE SAFE WORKING HYDROGEN PRESSURE


116

FIGURE 60 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR 1 OR 1 1/4% CR- 0.5% MO STEEL WITH 0.1

CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT EACH TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE CREEP STRENGTH GOVERNED DESIGN ALLOWABLE STRESS AND ON THE VERTICAL AXIS FOR THAT POINT AN

ESTIMATE OF THE SAFE WORKING HYDROGEN PRESSURE


117

FIGURE 61 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR ANNEALED 2 1/4% CR- 1% MO STEEL

WITH .03 CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT EACH TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE



118

FIGURE 62 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR ANNEALED 2 1/4% CR- 1% MO STEEL

WITH .06 CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT EACH TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE



119

FIGURE 63 – CURVES SHOWING FOR EACH TEMPERATURE THE HYDROGEN PRESSURE-EQUILIBRIUM METHANE PRESSURE RELATION FOR ACCELERATED COOLED 2 1/4% CR- 1% MO- .25V STEEL WITH .01 CARBON ACTIVITY. THE STRENGTH CURVE SHOWN INDICATES AT

EACH TEMPERATURE INTERSECTED THE METHANE PRESSURE NUMERICALLY EQUAL TO THE CREEP STRENGTH GOVERNED DESIGN ALLOWABLE STRESS AND ON THE VERTICAL AXIS

FOR THAT POINT AN ESTIMATE OF THE SAFE WORKING HYDROGEN PRESSURE


120

FIGURE 64 – PREDICTED SAFE OPERATING LIMITS FOR CARBON STEEL CONTAINING CEMENTITE (SEE FIGURE 58) BASED ON CREEP STRENGTH IN THE HIGHER TEMPERATURE

RANGE AND REACTION RATE IN THE LOWER TEMPERATURE RANGE


121

FIGURE 65 – PREDICTED SAFE OPERATING LIMITS FOR CARBON-0.5% MO STEEL CONTAINING CEMENTITE (SEE FIGURE 59) BASED ON CREEP STRENGTH IN THE HIGHER TEMPERATURE

RANGE AND REACTION RATE IN THE LOWER TEMPERATURE RANGE


122

FIGURE 66 – PREDICTED SAFE OPERATING LIMITS FOR 1 OR 1 ¼% CR- 0.5% MO STEEL CONTAINING STABLE CARBIDES (SEE FIGURE 60)BASED ON CREEP STRENGTH IN THE

HIGHER TEMPERATURE RANGE AND REACTION RATE IN THE LOWER TEMPERATURE RANGE


123

FIGURE 67 – PREDICTED SAFE OPERATING LIMITS FOR ACCELERATED COOLED 2 1/4% CR- 1% MO STEEL CONTAINING STABLE CARBIDES BASED ON CREEP STRENGTH IN THE HIGHER

TEMPERATURE RANGE AND REACTION RATE IN THE LOWER TEMPERATURE RANGE


124

APPENDIX A – BACKGROUND TO THE NELSON CURVES

PREPARED BY J.E. CANTWELL


125

1.0 HISTORY

The effects of hydrogen at high temperatures and pressures had been chronicled by C.A. Zapffe and studied by F.K. Naumann and others. There was general agreement that the attack involved hydrogen reacting with the carbides in the steel, forming methane. The methane cannot diffuse out, collects in the grain boundary area, building up pressure and causes enough fissuring that a stress overload failure eventually occurs. Naumann‟s and Bauklok‟s work supported the German desire for hydrogen processing to support oil and petrochemicals for their war effort. Much of Naumann‟s laboratory work was limited to short term testing. However, he did recognize that hydrogen attack was time dependent. In order to predict the effects of attack called for longer term laboratory work or collecting data from actual plant operation. Because of the war, there was no time to do extensive testing. G.A. Nelson of Shell Oil Development, began to collect available laboratory and plant operating experience. J. Schuyten of Shell Oil compiled an experience summary paper in 1947 based on a talk he gave in July of 1946. Schuyten summarized his hydrogen attack work as follows: 1) Hydrogen attack starts at a limiting temperature and partial pressure of hydrogen as a

function of time. The longer the exposure time, the lower the temperature and pressure limits. The higher the temperature, the lower the limiting pressure and, conversely the higher the pressure, the lower the limiting temperature. Short time tests are not adequate. See 190HFigure A-1.

2) The rate of hydrogen attack increases with increasing temperature and increasing pressure. 3) Hydrogen attack on steels is manifested by decarburization and intergranular fissuring of

the structure, which is reflected in reduced mechanical properties, particularly tensile strength, ductility, and impact strength. Under severe conditions, reduction of oxides and sulfides by the hydrogen also occurs. Reaction of the carbides may lead to blisters formed as a result of the pressures built up by the methane reaction product.

4) Hydrogen attack is progressive with time, although in many instances an induction period prior to chemical attack has been found.

5) Additions of alloying elements of the carbide forming type such as Cr, Mo, W, V, Ti, Nb, etc. substantially increase resistance to hydrogen attack. Noncarbide formers such as Ni, Cu, and Si are not effective. Low alloy steels such as C-0.5Mo, 1 and 2%Cr-Mo show good hydrogen resistance under moderately severe conditions. 3 and 6% Cr-Mo steels and 18Cr-8Ni stainless steels are used for conditions of greater severity.

6) Steels with high carbon content are more susceptible to hydrogen attack than steels of low carbon content.

7) Coarse grained steels are more susceptible to hydrogen attack than fine grained steels. 8) A sorbitic type structure has superior hydrogen resistance to a ferritic-pearlitic type

structure.


126

A paper by Nelson at API in 1949 covered Schuyten‟s summary and included both Allied and German experience from the 1939 to 1945 wartime period. The work presented hydrogen attack experience in the form of a chart plotting temperature versus hydrogen partial pressure, drawing a series of material specific curves through points of satisfactory operating experience. The curves were revised 22 times in the previous six years as a result of adding points of relevant experience. The solid lines represent limits of experience with an appropriate safety factor for internal decarburization, methane formation and fissuring . The dashed lines represent the competing surface decarburization reaction. See 191HFigure A-2. Important points brought out in the 1949 Nelson paper were:

1) In order to produce damage to the structure of carbon steel by hydrogen, both high

pressures and high temperatures are required. 2) Carbon steel varies in its resistance to attack with the manufacturing methods and , in

order to build commercial plants, limits for this material should be at a level where steel in its average condition will be adequate.

3) For protection against damage to steel by hydrogen, relatively small amounts of carbide stabilizing elements are required.

4) A chart showing limits based on operating experience for ordinary carbon steel and low alloy steels in contact with hydrogen was presented.

5) Equipment constructed with alloy liners should have provision for escape of hydrogen which diffuses through the liner.

6) When other corrosives are present in the process streams, measures for protection against those elements should be taken in addition to those required for resistance to hydrogen damage.

A 1958 Subcommittee on Corrosion survey of plant operating experience added a welded section to the curve for carbon steel. It was also concluded that “the limiting curves presented in Nelson‟s 1949 API paper were considered adequate.” See 192HFigure A-3. The Nelson Curves, as they became known, went through periodic revision as new laboratory and plant operating experience became available. Nelson‟s paper presented to API in 1965 offered a more complex set of curves modifying those from 1949. A welded or hot bent carbon steel curve was added. The C-0.5Mo curve was raised about 50oF. A set of curves covering trace amounts of molybdenum was added as well as curves showing time dependence for carbon and C-0.5Mo steels. The 1 Cr-0.5 Mo curve was changed to 1.25 Cr.-0.5 Mo. See 193HFigure A-4. Many of these changes were a result of API sponsored work at the University of Wisconsin in the 1960‟s directed by Dr. F.H. Vitovec. This research showed that:

The predominant mechanism in the hydrogen attack process is dependent on many variables including temperature, hydrogen pressure, time, alloy, heat treatment, surface condition and stress. For plain carbon steel, the combination of high temperatures and low hydrogen pressures seems to favor surface decarburization. The combination of low temperature and high hydrogen pressure favors internal attack and fissuring. Higher test temperatures increase the rate of carbide solution and carbon diffusion so that the surface reaction predominates. At lower temperatures the mobility of carbon is reduced, so that the reaction occurs at internal reaction sites. Hydrogen pressure has been shown to affect both the rate and extent of attack. High hydrogen pressures act to increase the hydrogen solubility in steel and provide a greater driving force for internal attack.


127

In the late 1960‟s, the API‟s Division of Refining‟s Subcommittee on Corrosion of the Committee on Refinery Equipment took over Nelson‟s work and in July 1970 produced API Publication RP 941 “Steels for Hydrogen Service at Elevated Temperatures and Pressures in Petroleum Refineries and Petrochemical Plants.” This document was basically Nelson‟s 1965 curves with 2.25 Cr-1.0 Mo and 0.25 molybdenum steels added along with sections on allowable stress, effect of heat treatment and effect of cold work. A curve was added for 1.0 Cr-0.5 Mo. Industry experience had shown blistering problems on Catalytic Reformer hot wall reactors so the curve was split off from the 1.25 Cr-0.5 Mo curve. See 194HFigure A-5. The second edition of RP 941 was published in June, 1977. The major revision involved lowering the curve for carbon-0.5 Mo steels based on attack reported by the industry. It also suggested that more frequent inspection of carbon-0.5 Mo equipment operating at conditions exceeding the revised curve was required. This edition also brought out that 2.25 Cr-1.0 Mo steels stress-rupture tested by MPC at conditions about 100oF above the curve at 3000 psi showed no impaired creep rupture strength but did suffer a loss of stress rupture ductility. The third edition of RP 941 was published in May, 1983. This edition reported more carbon-0.5 Mo attack instances and suggested more rigorous evaluation of carbon-0.5 Mo equipment particularly on catalytic reformers where most of the failures occurred. The fourth edition of RP 941 was published in April, 1990. The major changes were the addition of a section on liquid/vapor phase hydrocarbon systems with some hydrogen added and a continuing effort to define the carbon-0.5 Mo problem. Because of more industry information on C-0.5 Mo attack instances below the 1977 second edition C-0.5 Mo curve, C-0.5 Mo was removed from the main curves and presented separately. Cautions on the use of C-0.5 Mo steels were added along with suggestions calling for more rigorous inspection of existing C-0.5 Mo equipment. Use of various advanced ultrasonic techniques were suggested. Although the C-0.5 Mo curve had been lowered in 1977 and more frequent inspection suggested, many more questions on C-0.5 Mo seemed to come up after it was separated from the Figure 1 set of materials curves in the fourth edition. There are still many questions to be answered on C-0.5Mo steel‟s apparent variable resistance to hydrogen attack. A mechanism based on carbide stability as a result of heat treatment conditions may be at least a partial answer. (The Materials Properties Council has a group of sponsors for a program on C-0.5 Mo.) The fifth edition of RP 941 was published in January, 1997 and a supplement in April, 1998. Several more C-0.5 Mo points were added and a section on inspection procedures was included. Sections on hydrogen attack of 1.25 Cr-0.5 Mo and 2.25 Cr-1.0 Mo were added. See 195HFigure A-6 and 196HFigure A-7. The sixth edition published in March 2004 has some general wording changes, revised reference listing and revised Inspection section and tables.


128

2.0 WHAT DATA IS AVAILABLE?

Since the curves presented in API RP 941 are based mostly on operating data, there is always the question of “Can we come up with better data or validate the curve positions since we now have better data and tools to handle the data”? What is the data? There is a package that was given to API by Nelson and more that Nelson‟s widow passed on to API after his death. That package contained very little actual data on the individual points that were on the curves. Essentially, Nelson‟s 1949, 1958 and 1965 curves contain all the references and data available. The file contains some of the detailed information on C-0.5 Mo steel experience from the early 1970s when API took over the maintenance of the curves. The Shell files have been thoroughly researched for any supporting data several times in the last 35 years. None of Nelson‟s files have survived. Recent inquiries with Shell recently found nothing that pertained to Nelson‟s work. It was reported that any files that Nelson had when he retired were probably thrown out when Shell Emeryville was closed down in the early 1970‟s. There is considerable C-0.5 Mo experience in the form of correspondence and published technical papers from the early 1970‟s through the present. There is some correspondence and technical papers on 1.25 Cr-0.5 Mo and 2.25 Cr-1.0 Mo and some information on whether or not API RP 941 should be used in ammonia service where nitriding is a problem. Other than that, only the data that is on the original 1949, 1958, and 1965 curves or in the references is available. In order to present all of the available plant and laboratory data in a condensed form that is simpler to follow, the materials Properties Council maintains a summary spreadsheet. The spreadsheet has 48 available information columns and 835 entries. Some data from the NACE T-8 minutes from 1957-1996 Refin-Cor 3.0, the 941and other available references back to the 1930‟s was added. Japanese data in the form of two separate spreadsheets covering C-0.5Mo and 2.25Cr-1Mo data is included. The reference list for the spreadsheet is attached as Attachment 1.


129

3.0 WHERE DID THE DATA COME FROM?

The original curves in the 1949 Nelson paper were generated by connecting satisfactory plant data points and plotting unsatisfactory laboratory data. There was a safety factor included in the curves. In discussions, the safety factor seems to be on the order of roughly 30oF for the carbon steel curve where apparently most of the early data was available. Some refiner‟s add a safety factor of from 30oF to 50oF to the curves so that in effect a conservative factor of as much as 80oF might be used when carbon steel is considered for hydrogen service. The low alloy curves were drawn with much less experience and therefore fewer data points and unknown safety factors. The 1949 Nelson paper states that the carbon steel data at >2600 psi hydrogen partial pressure came from Standard Oil Development Company and Hercules Powder. The carbon steel data at <2600 psi hydrogen partial pressure came from Shell, M.W. Kellogg, Standard Oil of California, Zapffe and Evans (DuPont). The C-Mo data came from hydrogenation plants built during World War II with data just becoming available when the paper was given at API. The low Cr-Mo data - .i.e. 1Cr-0.5Mo was actually from Cr-V steels used successfully for hydrogenation vessels during the war. There was considerable Cr-Mo used for piping. It was assumed that Cr-V and Cr-Mo steels had approximately equal hydrogen attack resistance. There were very few 2Cr-Mo data points. The data for the 3Cr-Mo and 6Cr-Mo steels go to 6000 psi hydrogen partial pressure. German experience from World War II showed that these steels needed to be fortified with carbide formers such as W and V to make them hydrogen attack resistant at very high pressures. The research also showed that the strong carbide formers Ti, W, Mo and V helped impart hydrogen attack resistance at very high pressures. However, because of forming and/or fabrication problems only significant amounts of Mo were used for commercial materials. The following is a summary of all the data points on the Fifth Edition of API RP 941 1997/1998 curves. The summary includes the source of the information, the exposure conditions reported and the available data. Attack is defined as surface decarburization or internal decarburization and fissuring or cracking.

3.1 API Publication 941 Fifth Edition Supplement 1 April 1998

Figure 1 Points Point 1. Information from Shell. No correspondence Point 2. Information from Timken Roller Bearing. No correspondence Point 3. Paper available by F.K. Naumann. Point 4. Paper available by N.P. Inglis and W. Andrews. Point 5. Paper available by J.L. Cox. Point 6. Paper available by R.J. Sargant and T.H. Middleham. Point 7. Information from Standard of California. No correspondence Point 8. Information from DuPont. No correspondence Point 9. Information from Ammoniawerk Merseberg. No correspondence Point 10. Information from Hercules Powder. No correspondence Point 11. Paper available by C.A. Zapffe.


130

Point 12. Information from M.W. Kellogg. No correspondence Point 13. German operating experience. No correspondence Point 14. Information from Vanadium Corporation of America. No correspondence Point 15. Information from Imperial Chemical Industries. No correspondence Point 16. Paper available by T.C. Evans. Point 17. Information from Norweg Hydroelectric. No correspondence Point 18. Information from Union Oil. See 18T. Correspondence available. Naphtha hydrotreater carbon steel reactor effluent line showed scaling but not high temperature hydrogen attack after 18 years at 500F and 504 psi hydrogen partial pressure. Point 19. Paper available. Information from Exxon. Considerable data. Attack very spotty. Difficult to match attack with temperatures. Temperatures varied considerably. Attack propagated faster than anticipated. Problem on carbon steel Catalytic Reformer reactors (cold shell) due to bypassing behind the refractory. Attack went through wall in about a year. Decarburization, fissuring and cracking which caused an in-service leak. Point 20. Information from an API survey. Correspondence available. Four carbon steel points added. Two showed decarburization, one attack and one unsatisfactory. Information from Amoco, Standard of California, Union Oil and Exxon. Point 21. Information from Air Products. No correspondence Point 22. No paper available. By G.D. Gardner and J.T. Donovan. Point 23. Information from Amoco. No correspondence Point 24. No paper available. By E.W. Comings Point 25. Paper available by M. Hasegawa and S. Fujinaga. Point 26. No correspondence Point 33. No correspondence

3.2 Figure 1 Points C-0.5Mo Appendix A Point 18. Information from Unocal. Very little information. Ammonia Plant Syngas cooler/ Boiler Feedwater heater. Tubes attacked only in the tubesheet. Temperature 625oF, hydrogen partial pressure 1712 psi. Original micros. Slight decarburization and fissuring. Point 27. Information from Unocal. Considerable information. Catalytic Reformer reactor to interheater line ruptured resulting in a major fire. Several other lines showed attack. Very similar to point 36. Spotty attack with blisters, decarburization and fissuring. Felt the attack was only in areas that had a susceptible microstructure. Considerable temperature data. Typical conditions were 860oF and 370-400 hydrogen partial pressure. Good copies of micros. In addition to the failure, hydrogen attack was also noted in the shell of the hot combined feed exchanger and the line to the Charge heater. In 1971 three years before the failure, after 1.5 years service a C-0.5Mo flange on the failed line was found to be hydrogen attacked but the elbow attached to it was not attacked. There were blisters decarburization and fissuring in some areas and laminations with no decarburization in other areas. A February 11, 1976 letter outlined 21 experience cases with only two cases of high temperature hydrogen attack.


131

Point 28. Information from Amoco. Two incidents. Catalytic Reformer with swing reactors. First incident. Temperature 850oF typical, 880oF maximum; hydrogen partial pressure 240 psi. After two years had multiple piping leaks. Transverse weld cracks into base metal. No postweld heat treatment. Felt that weld defects created high stress areas. Second incident. Temperature 880-930oF, hydrogen partial pressure 195 psi. Catalytic Reformer exchanger developed cracks three times in 10 months. Was in service 16.5 years before upstream exchanger was removed. Cracking found after two years additional service. Cracks started in shell to flange weld and extended into the base metal. Not much information. A third incident was reported on a buckslip attached to letter of February 4, 1976. Temperature 850-860oF, hydrogen partial pressure 405-415 psi. Failed in seven years. This information was not plotted. Point 32. Information from Citgo. Ammonia Plant. Very little information. Normal temperature to condenser 400oF, hydrogen partial pressure 4000 psi. Bypassing in converter increased temperature to about 480oF to the water box type condenser during the last year of operation. Harp connector tube ruptured. Severe decarburization and fissuring. Temperature may be questionable since it was estimated. Good micros. Point 34. Information from Koch. Desulfurizer reactor effluent/hydrogen preheat exchanger shellside. Temperature 720oF, hydrogen partial pressure 840 psi. Temperature and hydrogen partial pressure may be questionable since they were estimated. Attack (decarburization/fissuring) noted at the shell inlet and outlet nozzles. Noticed blisters during inspection. Micros are copies. Less damage on the cold end of the shell although voids seen. Point 35. Point not used. Point 36. Information from Texaco. Catalytic Reformer (Platformer). Temperature 778oF average, 806oF maximum; hydrogen partial pressure 345 psi Good temperature data and micros. Combined feed exchanger to the Charge heater piping. Leaked in service with very minor fire noted after 80,000 hours service. Considerable data available. Spotty decarburization and fissuring through the wall with good material on either side. Noted helical pattern of blisters on the inside. Supposed to have been postweld heat treated. A similar case at a sister refinery that involved the hot combined feed exchanger shell as well as piping to the Charge heater. Temperature 720-830oF; hydrogen partial pressure 330 psi. Only one side of the parallel exchanger train suffered problem. Two shells in the same position suffered severe blistering but no decarburization or fissuring. Each had seven years service. Extensive temperature profiling showed that there was up to 150F difference in temperature between the two trains depending on the feed rate. Extensive temperature data available. The second case was not reported to API. Similar to point 27. Not near welds in either case. Point 37. Information from Exxon. Swing reactor Catalytic Reformer reactor effluent piping. Cracking found by radiography. Almost through wall. In heat affected zone and base metal at least partially due to hydrogen attack. Other reasons not explained. Only had problem on non postweld heat treated piping. Showed decarburization and fissuring. Good micros. There was some blistering on fittings but no decarburization and fissuring. Data was reviewed and deemed representative. Operated about one year over the curve but felt they were over the curve less than the incubation time.


132

Point 38. Information from Exxon. Swing reactor Catalytic Reformer reactor effluent piping. Cracking found by radiography. Very spotty localized decarburization and fissuring in the heat affected zones and base metal. Copies of micros. 9.5 years service. Temperature 620-675oF, hydrogen partial pressure 270 psi. Temperature is low with no explanation. No postweld heat treatment. Point 39. Information from Exxon. Swing reactor Catalytic Reformer reactor effluent piping leaked at weld. Localized decarburization and fissuring in weld metal. Temperature 698-752oF, hydrogen partial pressure 285-300 psi. In service about 12 years with about one year at maximum temperature. Point was plotted at average temperature. Copies of micros. Was postweld heat treated at 1250oF. Point 41. Information from Caltex. Catalytic Reformer (Platformer) Combined feed exchanger. Shell outlet flange cracked. Found during inspection. Complex information. Very difficult to estimate temperatures on these shells. Temperature 748oF average, 770oF maximum; hydrogen partial pressure 349 psi average, 417 psi maximum. About 16,000 hours at maximum temperature, 1600 hours at maximum partial pressure. Total service time 86,000 hours. Plotted at maximum estimated temperature/average hydrogen partial pressure. Decarburization and fissuring. Included several samples from cracked shell welds. Point 42/43. Information from Getty. Catalytic Reformer (Platformer) Combined feed exchanger channel. Long time multi location service. Channel inlet(hot) has significant attack with decarburization and fissuring(42). The channel outlet(cold) has very slight decarburization and fissuring. Considerable information and committee discussion over temperature. Channel inlet 750oF, channel outlet 623oF for the last nine years; hydrogen partial pressure 350 psi. Channel temperature was about 100oF cooler for the first 12 years of service. Attack found during inspection. Point 44. Information from Caltex. Catalytic Reformer (Platformer) Combined feed exchanger shell. Decarburization and fissuring found after shell leaked during hydrotest. Attack was in the base metal not near any welds. Temperature 735oF, hydrogen partial pressure 313 psi. Little temperature data but numbers look ok. Exchanger was in series (small unit). Micros are copies. Attack was on the shell under the outlet nozzle in two or three areas. Short term carbon steel replacement showed no detectable attack after about 1000 hours service at the same operating conditions. Point 45. Information from Japan Steel Works. Hydrodesulfurizer exchanger channel. 640oF maximum temperature, 620oF average; 457 psi hydrogen partial pressure. Nine years operation. Channel had blisters, decarburization, fissuring and voids in the base metal to about half of 30 mm wall. Found by inspection. Point 46. Information from Japan Steel Works/Japan Gasoline Corporation from an Exxon affiliate. Swing reactor Catalytic Reformer reactor effluent piping cracking found by radiography. In field weld which was postweld heat treated. In weld metal and heat affected zones. Temperatures were plotted about mid range between maximum and average. Temperature 626-680oF, hydrogen partial pressure 350 psi. Cracking but no decarburization in the weld and heat affected zones. No base metal cracking or decarburization but some minor fissuring at inside near weld..


133

Point 47. Information from Japan Steel Works. Gas Oil Hydrodesulfurizer piping elbow. STPA12 had sporadic cracking between weld metal and heat affected zone. Decarburization, fissures, cracks and voids after seven years service. Was not postweld heat treated. Found by inspection. Temperature 684oF average, hydrogen partial pressure 738 psi. Point 48. Information from Air Products. Ammonia Plant Converter exit piping cracked in base metal banded area in microstructure about 45% of wall. Temperature was 550oF for the first 7 years of service. The temperature the next 9 years averaged 570oF. The average temperature for the last two years was 670oF which was about +100oF over the plotted point, which brings the point almost up to the 1977 curve. Good original micros show decarburization and fissuring in the banded structure. Good information. Point 49. Information from Texaco(Getty). Strange case. Desulfurizer reactor weld that were backclad with type 309. Attack only in the welds in this 4 7/16 inch thick reactor. Looks like high temperature hydrogen attack. Decarburization and fissuring in weld showed up as longitudinal cracking. Considerable transverse weld cracking probably due to fabrication problems. No attack in base metal under spot welded type 405 liner. Similar problem 10 years earlier with two other desulfurizer reactors. Considerable information and disagreement in the committee on these three incidents. Good micros. 28 years service. Temperature 750-770oF, hydrogen partial pressure 390 psi. Point 50. Information from Mobil. Hydroprocessing unit reactor feed/effluent exchanger on or higher than the curve. Temperature was 575oF, hydrogen partial pressure 1200 psi for the first 10 years service. The next three years the temperature was 630-670oF and the last 4 years the temperature was 670-730oF. Tubesheet forging found cracked between ligaments and tubes blistered with no decarburization and minor fissuring after 17 years total service. Some decarburization and fissuring with voids. Good micros. Little information. Repairs were not successful. Point 51. Information from Shell. Desulfurizer reactor effluent/ hydrogen heater. Could be plotted slightly higher. Attack was under type 304 clad that had sensitized and intergranularly cracked for several years. Temperature was 690oF average, 725oF maximum; hydrogen partial pressure was 397 psi average, 500 psi maximum. Base metal decarburization and fissuring were found after about 20 years service. Micros are copies. Point 52. Information from Texaco. Hydrocracker reactor feed/effluent exchanger shell. Operations bled recycle hydrogen into system to minimize feed side fouling problem. Found during shutdown when could not pull bundle. Temperature 680oF, hydrogen partial pressure 1725 psi. Type 405 rollbond clad was severely blistered. Severe fissuring in base metal found after opening blister. Fissuring almost went through shell wall. Two phase flow situation. Original micros. Little data.


134

Point 53. Information from Kemira. Ammonia Plant Syngas Loop 24 inch forged pipe. Temperature 545oF, hydrogen partial pressure 2190 psi. Ultrasonics eventually detected fissuring 46 mm through 76 mm wall. Micros detected fissuring/voids through 64 mm of the wall. Only had attack on the forged pipe. Hot rolled pipe in the same circuit showed no detectable attack. Followed propagation with ultrasonics. Went much faster than was expected. Micros are copies. Severe decarburization and fissuring. Felt problem was due to susceptible microstructure due to difference in tempering temperature. No data on operating temperatures. Point 54. Information from Chevron. Catalytic Reformer Feed/Effluent exchangers. Inspection found surface blisters on flange with base metal decarburization and fissuring. About 100oF below curve. Channel nozzles on C and D exchangers which were the cold two exchangers. Operated at about 750oF and 300 psi hydrogen partial pressure. Could well have been higher temperature due to possible interchangeability and parallel trains. Attack was to about 0.4 inches in a banded microstructure. No decarburization. High stress area. Point 55. Information from Chevron. Catalytic Reformer large diameter piping. Leaked in service, small fire. Heat affected zone and base metal fissuring about 100F below curve. Temperature 800oF average, 850oF maximum. Hydrogen partial pressure was 175-190 psi. No postweld heat treatment. Crack in elbow near Charge heater. Cracking in same location as point 58. Microstructure was different probably due to hot bending. Point 56. Information from Chevron. Catalytic Reformer exchanger just below curve. Inspection found surface blisters with decarburization and fissures in the base metal. Postweld heat treated. Hydrogen partial pressure was 275-300 psi. Temperature 825oF maximum, 810oF average. Operated above the curve for a short time. Shallow decarburization and fissuring at blisters and inclusions. Both hot shells in the parallel train were blistered. No problems noted on the cold shells. Point 57. Information from Chevron. Catalytic Reformer large diameter piping showed decarburization and fissuring in the weld/heat affected zone and base metal. (doesn‟t show on curve). Inspection found two of 28 circumferential seams in the piping between the feed/effluent exchangers (parallel train) and the Charge heater were found cracked. Weld number 8 had one circumferential and one transverse crack. Weld number 25 had three transverse through wall cracks. The crack in #8 weld was between the heat affected zone and the fusion line and showed decarburization and fissuring. Temperature 850oF maximum; hydrogen partial pressure 225 psi maximum. Point 58. Information from Chevron. Catalytic Reformer large diameter piping double submerged arc welded bent pipe elbow showed heat affected zone decarburization and fissuring near 1977 curve although plotted pretty low. Cracking in elbow in line from the feed/effluent exchangers to the Charge heater. Elbow at the heater. Same location as point 55, different refinery. Temperature was 855oF maximum, 810oF average and the hydrogen partial pressure was 170 psi maximum. 20 inch diameter, 0.5 inch wall. Circumferential weld crack which leaked was found by radiography next to the heat affected zone on the elbow side and was decarburized and fissured. No postweld heat treatment.


135

Point 59. Information from Chevron. Hydroprocessing unit piping showed decarburization and fissuring in base metal. Plotted ave conditions. Maximum would be at curve. Was postweld heat treated. Data sheet only. Temperature 550oF average; 600oF maximum. Hydrogen partial pressure 1800-2200 psi. Point 60. Information from Chevron. Hydroprocessing unit exchanger channel forging showed decarburization and fissuring in the base metal. Was postweld heat treated. Could have been plotted close to the curve. Temperature 550oF average; 600oF maximum. Hydrogen partial pressure 1800-2200 psi. Point 61. Information from Chevron. Hydroprocessing unit piping showed base metal decarburization and fissuring. Plotted average conditions. Maximum conditions could be at or higher than the curve. Data sheet only. Was postweld heat treated. Temperature 530oF average; 600oF maximum. Hydrogen partial pressure 2100-2300 psi. Point 62. Information from Chevron. Naphtha Hydrotreater exchanger (shell). Showed some surface blisters and heat affected zone and base metal decarburization and fissuring. Temperature is very low. Data sheet only. Temperature 670oF average; 700oF maximum. Hydrogen partial pressure 190 psi. Point 63/64. Information from Tosco. Hydrotreater reactors with rollbond type 405 clad that had blistered. Blisters found during inspection after 13 years service. Attack (decarburization and fissuring) found 14 years later by boat sampling. Plate was 3-5/16 inch wall. Micros not furnished; Harwell report has good ones. Temperatures back seven years were pretty low. However, could have been higher previously. Point 63 temperature 600oF average; 750oF maximum. Hydrogen partial pressure 500 psi. Point 64 temperature 600oF average; 770oF maximum. Hydrogen partial pressure 525 psi. Considerable information in Harwell report on inspection methods evaluation to find high temperature hydrogen attack. One of the reactors plus other companies samples. Number 64 from the Harwell report had decarburization and fissuring up to 12.2 mm. Microstructure showed considerable inclusions and segregation with decarburization under the blisters. Another sample from the number 64 reactor showed hydrogen attack up to 37 mm pointing up the spotty nature of the base metal attack. Plate had considerable inclusions which probably affected earlier ultrasonic examinations. Seemed to be a layer of decarburization under areas of thin bond layer. Microstructure was banded. Point 65. Information from Exxon. Hydroprocessing unit reactor inlet self-reinforced nozzle decarburized and fissured under type 309 overlay only in thick sections. Overlay blisters found during inspection.. Ultrasonics showed considerable cracking in heavy sections of nozzle. Temperature 775F, hydrogen partial pressure 550 psi. Data showed that attack was about on the curve. Gets into heavy wall diffusion since fissuring was only in the very heavy sections. See Battelle report on hydrogen in base metal under austenitic clad/overlay.

3.3 File Correspondence but Points Not Plotted

1. Exxon letter of May 8, 1985. Liquid phase hydrogen attack showing decarburization and fissuring of carbon steel in a nozzle and line from a separator. Failed in service. All the attack was in high stress areas. Conditions were 650oF and 497 psi hydrogen partial pressure in service 15 years.


136

2. Gulf letter of March 23, 1976 gave data on three pieces of C-0.5Mo equipment (two reactors and a line) that had not had any hydrogen attack problems. 3. Exxon letter of April 6, 1976 showing SoCal, Union Oil, Amoco and Exxon experience for the discussion on lowering the C-0.5Mo curve which was done in the second edition of API RP 941 in 1977. Showed both attack and no attack points. 4. Arco Catalytic Reformer reactor effluent line (24 inch) had a mistakenly welded carbon steel pup piece in. Failed in service. Showed extensive decarburization and fissuring after being exposed for 4060 days at average temperature of 810oF and 840 days at maximum temperature 950oF and 160-190 psi hydrogen partial pressure. 5. Trinidad, W.I. refinery Catalytic Reformer hot wall A301 grade B reactor with about 300,000 service hours showed some voids with decarburization and fissuring near blisters. Found during inspection. Some “Creep embrittlement” problems also noted. Operating conditions were temperatures 882-930oF and hydrogen partial pressures of 300-410 psi. 6. Union Oil report showed Catalytic Reformer hot wall A301 Grade B reactor had blistering in one area with decarburization but no fissuring. Inspection found that about 20% of a 10 foot by 10 foot area was blistered. 7. PEC Rhin in France letter on an Ammonia Plant Sygas converter exit line that leaked after 3000 days of service at 545oF and 15 MPa hydrogen partial pressure. Material was 20CrMoV13.5.(3Cr-0.5Mo-V) Werkstoff Number 1.7779. The line had a history of cracking problems and high hardness. Some nitriding. Line was postweld heat treated below 1350oF. Showed decarburization and fissuring. 8. Information from Braun, BASF and others. Considerable correspondence on Ammonia Plant hydrogen attack possibilities, hydrogen cracking and nitriding. 9. Information from Texaco not submitted. Catalytic Reformer Naphtha Hydrotreater reactor feed/Platformer effluent exchanger channel nozzle failed after 12 years service. Was carbon steel not the specified C-0.5Mo. Cracked, decarburized and fissured. Temperature 600oF ; hydrogen partial pressure 460 psi. 10. ICI in 1976 submitted several C-0.5Mo incidents through Kellogg. Had blistering on two Ammonia Plant Shift Converters but no problems on a third. Showed blistering, no decarburization but fissuring to 10 mm of 90 mm walls. Good micros. Temperature 784 - 813oF and hydrogen partial pressure of 165 psi. Comment made that they had several C-0.5Mo line failures in welds. Also mentioned they had noted high temperature hydrogen attack on several Aromatics unit reactor transfer lines(Catalytic Reformer) in welds that were made with carbon steel on C-0.5Mo piping. Welds were about 0.2% Mo in service 2 years at 734oF and 210 psi hydrogen partial pressure. Data not plotted. 11. DuPont letter of December 12, 1980. Submitted four data sheets but no photomicrographs.


137

A. Carbon steel A106 pipe ruptured after 450 days at 610-630oF and 3600 psi hydrogen partial pressure. Attack was assumed to be decarburization and fissuring through about 90% of pipe wall. B. A105 carbon steel forged hub had about 0.010 inches of decarburization and fissuring in the base metal but through wall in the heat affected zone at 610-630oF and 3600 psi hydrogen partial pressure. C. Figure 1S is a 1-1/4Cr-1/2Mo valve body that showed no attack in 450 days. D. C-0.5Mo pipe welded with carbon steel electrodes. Temperature 660oF and hydrogen partial pressure 182 psi. In service between 1460-3830 days. 12 samples no attack noted. Mo ranged from 0.25% in the cover pass to 0.50% in the root. 12. Kellogg reported good experience with C-0.5Mo at 825oF and 185 psi hydrogen partial pressure in Ammonia Plant Shift Converter service for up to 15 years. 13. An Ammonia Plant in the Netherlands. C-0.5Mo weldments in a line in Converter feed service showed high temperature hydrogen attack after 11 years at 295C and 158-162 bar hydrogen partial pressure. Cracks were found by ultrasonics in the apparent heat affected zones of several welds between pipe and elbows. Assemblies were apparently normalized after welding since no heat affected zones were apparent. Decarburization and fissuring developed into cracks at the fusion line. The elbows were 52 mm wall. 14. Koa Catalytic Reformer hot wall reactor outlet nozzle. C-0.5Mo nozzle welded into a 1-1/4Cr-1/2Mo reactor bottom head with l-1/4Cr-1/2Mo electrodes. Longtime service at 950-975oF and 200psi hydrogen partial pressure. Attack on bore of heavy self reinforced nozzle to about 10-12 mm in 146 mm thick area and less in 51 mm area. Showed decarburization and fissuring. 15. VEB Schwedt sent a letter in 1989 giving an experience they had with 16Mo5 (C-0.5Mo) in an Ammonia Plant Desulfurizer Guard Case. They had a major blistering problem at an average temperature of 662oF and 247 psi hydrogen partial pressure after 115,000 hours service. Showed decarburization and fissuring of the weld metal. Blamed problem on a low heat treating temperature forming wrong carbides. 16. Unocal letter of July 6, 1987 described two cases of no attack on C-0.5Mo after several years service. A. Catalytic Reformer Naphtha Hydrotreater heater feed line (C-0.5Mo) at 555oF and 450 psi hydrogen partial pressure showed no attack after 25 years service. B. Catalytic Reformer Naphtha Hydrotreater reactors (C-0.5Mo) showed no attack after 16 years at 580oF and 330/270 psi hydrogen partial pressure. C. An 18 inch A335 P1(C-0.5Mo) line on the same unit operating at 785oF and 390 psi hydrogen partial pressure for 16 years showed no hydrogen attack but slight decarburization and some laminations.


138

17. PDVSA Curacao Catalytic Reformer Combined feed exchanger during wet fluorescent magnetic particle inspection showed weld cracking in the shell head cover weld to the effluent piping after 27 years at 750-783oF and 233 psi hydrogen partial pressure. Good micros taken from replicas. Cracking was in the weld/heat affected zone. Showed a bainite microstructure in the coarse grain heat affected zone. Could not tell if decarburization but was fissured. Looks like the temperature could have been closer to the reactor effluent than the feed outlet. 18. Mobil submission of 1996. Little detail but probably from a Hydrocracker. C-0.5Mo pipe had some fissuring and decarburization in the base metal but was not plotted because of questions on micros. Service at 595oF and 2065 psi hydrogen partial pressure for 9800 days. Found by ultrasonics. 19. Hydrogen Service Failures in Japan by Yokogawa. A carbon steel line under a relief valve on a heavy oil desulfurizer failed after 12 years service at 11 MPa hydrogen partial pressure and >473K. Pipe was insulated. Was decarburized and fissured through the wall. Ruptured March 31, 1982 resulting in a large fire and five fatalities. 20. Conoco submitted a carbon steel Catalytic Reformer butterfly control valve body that was in service for 10,800 hours at 675-750oF (715oF average) and 270-240 psi hydrogen partial pressure (240 psi average). Showed no attack. Very low Cr/Mo residuals. 21. Sohio submitted at letter October 10, 1961 with some detail on 11 samples. Five showed decarburization on carbon steel and one showed fissuring with analyses figuring Mo equivalents. A 1Cr-1/2Mo Catalytic Reformer reactor showed blisters, decarburization and some fissuring. 22. Information from Texaco, not submitted. Catalytic Reformer combined feed exchanger channel. C-0.5Mo unknown service life. Temperature 988oF maximum, hydrogen partial pressure 182 psi. Visual inspection and boat sample showed no evidence of hydrogen attack.

3.4 Points Plotted for 1-1/4Cr-1/2Mo and 2-1/4Cr-1Mo Point 1 Appendix B. Information from Chiyoda. Case A was a 1-1/4Cr-1/2Mo small bore pressure gauge nozzle in a Catalytic Reformer reactor effluent piping. The nozzle broke off during a shutdown after 26 years service. The nozzle weld was postweld heat treated in the field during construction. Operating conditions were 959oF average and 990oF maximum temperature and a hydrogen partial pressure of 331 psi. The schedule 80 1-1/2 inch nozzle was JIS STPA 23. The failure was in the heat affected zone and may have been complicated by “creep embrittlement”. The microstructure showed surface and internal decarburization, fissuring intergranular cracking and voids in the heat affected zone and base metal. Case B was a 12 inch Catalytic Reformer reactor effluent line that operated at an average temperature of 977F and a hydrogen partial pressure of 354 psi for 14 years. The material was JIS STPA 23. Blistering on the inside of an elbow was noted during a shutdown after 13 years service. The elbow was removed after about one year‟s additional service. Surface and internal decarburization, blistering, intergranular cracking , fissuring and voids were noted in the base metal. Considerable number of inclusions were noted . The chromium content of the piping in both cases was felt to be on the low side of the 1-1/4Cr-1/2Mo specification at 1.09 and 1.10%.


139

Point 2 Appendix C. Information from Exxon. A 2-1/4Cr-1Mo mixing tee from a Hydrocracking unit leaked in service. The tee mixed hot recycle and cold makeup hydrogen streams ahead of the No. 1 reactor. There was considerable thermal fatigue cracking in the tee along with some intergranular cracking at the crack tips. Some internal decarburization, fissuring and voids were noted at the inside surface of the tee as well as to about 4 mm deep on a sample of downstream piping. The operating conditions varied from about 850oF to 940oF at a hydrogen partial pressure of about 1410 to 1590 psi for the hot hydrogen stream and about 580oF to 810oF at a hydrogen partial pressure of about 1200 to1550 psi for the mixing tee and downstream piping. No hydrogen attack was noted in samples from the hot hydrogen stream. It was felt that the apparent high thermal stresses at the tee may have contributed to the hydrogen attack problem in the tee and the downstream piping. The piping detail had been in service about 20 years.

3.5 Attack / No Attack Points The fifth Edition of API Publication 941 shows the following attack/no attack point distribution on the curves:

Carbon Steel 21 attacked 30 not attacked

C-0.5Mo 37 attacked 31 not attacked

1 or 1-1/4Cr 6 attacked 11 not attacked


3Cr 4 attacked 3 not attacked


Nelson‟s original 1949 curves had the following point distribution:

Carbon Steel 12 attacked 18 not attacked

C-0.5Mo 3 attacked 4 not attacked





Of the data making up the curves, it is noted that while 73 attack points are shown in the API RP 941 1997/1998 curves, only six actual failures are noted.

The rest of the attack incidents were found before failure by various inspection technique‟s including visual, ultrasonics, radiography, magnetic particle inspection, hydrotest and shutdown maintenance. Some of the failures were due to wrong materials or specification break problems such as bypass lines


140

A review of the Japanese C-0.5Mo data indicates six attack instances in the approximately 120 surveyed case histories. A review of all the known plant data points in the 941 package showed that there were 25 instances of attack on carbon steel with two known leakers and two complete in service failures. With C-0.5Mo there were 66 instances of attack with five leakers and five complete in service failures plus one graphitization failure that is not on the curves. With the Cr-Mo steels there were 15 instances of attack with two known leakers. Many of the leakers caused relatively minor fire damage. Most of the complete failures caused major fire damage and some involved fatalities. A total of 106 attack points in the 941 package showed nine leakers and eight complete failures. There are probably more leakers and complete failures but the attack incident information was not detailed enough to draw any conclusions. There obviously have been attack incidents and complete failures that have not been reported to API. However, considering the amount of equipment and piping in high temperature hydrogen service in the industry, the percentage of attack problems is very small and the number of leakers/complete failures is much smaller than that.


141

4.0 DISCUSSION

A review of the data from the Hydrogen Attack Panel files leads to the conclusion that it is very difficult to use the plant data to verify the 941 curves. A major problem is that there is little or no supporting correspondence for the data points or more precisely how the curves were drawn from the early 1940‟s up until when the attack problem with C-0.5 Mo became generally known. Most of the C-0.5 Mo data became available in the middle 1970‟s to the middle 1990‟s. Even with the C-0.5Mo information, the lack of specific details makes any verifying exercise very difficult at best. Typically, the alloy composition is not detailed enough or the temperature or hydrogen partial pressure is a wide range which makes it difficult to identify limits with precision or analyze results. For instance, a typical semi-regenerative Catalytic Reforming unit may have a start of run to end of run reactor temperature range of 100oF or more. Parallel exchanger trains may have at least that much difference in outlet temperature depending on piping arrangement and charge rate. Three of the best data point sources for looking at the details of a specific attack instance are the carbon steel data from Point 19 and the C-0.5 Mo data from Point 27 and Point 36. However, all three have a wide spread in operating temperatures. Another problem is that there is no agreed on method to handle varying conditions. The stated approach by API has been to gather attack data as well as successful experience. However, in practice, more detail is provided for the attack instances. A review of all the available data and correspondence in detail and leads to questions about the location of some of the lower C-0.5 Mo points on the curves. Most of the materials‟ curves have stayed pretty much in the same general location over the years except for the C-0.5Mo curve which started out with a fairly high no attack temperature in 1949 below about 750 psi hydrogen partial pressure but was lowered significantly with the benchmark 1977 curve. In fact, the curve was raised about 50oF in the mid 1960‟s due to good experience. However, there was considerable disagreement in the subcommittee when the 1977 curve was lowered. Much of the argument was that there was considerable equipment in service that showed no evidence of hydrogen attack problems while operating above the new curve for several years. Several people recognized that lowering the curve would create a major inspection problem. 197HFigure A-8 shows the 1949, 1958, 1962, 1977 and 1997 C-0.5Mo curves. As was pointed out in several papers by Nelson, the C-0.5Mo and the Cr-Mo curves were drawn with considerably less data than the carbon steel curve. As more attack data has been accumulated the location of the C-0.5Mo curve came into question.

There was considerably more data with which to draw the carbon steel curve including safety factor of about 30oF. Adding to the safety factor was the fairly common practice of adding 35oF to 50oF to the assumed material design conditions. This could amount to as much as 80oF conservatism and would help explain why there have been few hydrogen attack problems with carbon steel unless the design temperature has been grossly exceeded. This could happen where the operating temperature was assumed to be considerably less than actual in a bypass line or a poor (in retrospect) specification break location.


142

As a result of the general lack of specific data, the tendency has been to emphasize 941‟s problem areas. This has been particularly true since the C-0.5 Mo attack problems have become apparent. In general, the curves have served the industry well over the roughly fifty years they have been available. One problem is with C-0.5 Mo, where carbide stabilizing alloying microstructure elements are fairly lean and heat treatment may have a considerable affect on attack. Other problem areas seem to be:

Where the materials are being affected by creep or are operating near or in the creep

range. Using the wrong material for the service conditions or the designers not doing a good job of

predicting the exposure conditions. Quantifying varying temperature, hydrogen partial pressure and time effects is a problem.

There is very clearly a problem on deciding where to plot a single point from data that includes a range of temperatures over time. There is a need for a parameter that would account for varying temperature and time similar to a tempering parameter for steels used to predict strength or creep/stress rupture. The only really reliable way to get an equipment exposure temperature is to properly measure the actual temperature of the component. Many times the process thermocouples are not well located for measuring the temperature of a particular component or in the case of exchangers, a particular exchanger in a multi-exchanger train. The design and/or process flow diagrams may not provide a very good estimate of actual operating temperatures. The location of some of the plotted points below the 1977 C-0.5 Mo curve might be questioned on this basis and the range of reported operating temperatures.

The increasingly complicated material microstructure with added alloying elements.

Explanations for C-0.5 Mo‟s apparent variable attack resistance have ranged from the type of carbides formed during fabrication/operation, to some effect on actual hydrogen partial pressure seen by the C–0.5Mo because of being in a sulfiding service or protected by a stainless steel clad or weld overlay. API sponsored projects at Battelle in the early 1980‟s showed that a sulfide scale could lower the effective hydrogen partial pressure seen by the backing material. Work at Battelle showed that austenitic stainless steel clad or weld overlay will lower the effective hydrogen partial pressure of the backing material. In a hydrogen / hydrogen sulfide environment, C-0.5 Mo does not have adequate corrosion resistance because it does not form a good, protective sulfide scale. Some C-0.5 Mo equipment and piping might have been put in hydrogen service where it was felt that the corrosion /scaling rate was manageable but generally, C-0.5 Mo is used in noncorrosive service. The work at Battelle on clad and weld overlay through MPC for API in 1984 presented a method of calculating the effective hydrogen partial pressure of the backing material under austenitic stainless steel clad or weld overlay. (See Appendix D) Some refiners use this method as part of their program of prioritizing C-0.5 Mo equipment for hydrogen attack inspection. There is one data point that shows blister attack on C-0.5 Mo austenitic weld overlay only in the very thick sections of a self reinforced nozzle. Currently, many designers are not giving C-0.5 Mo any credit over carbon steel in hydrogen service. This is probably overly conservative but is certainly on the safe side. However, the industry is faced with a difficult inspection/risk prioritization problem.


143

The notion of incubation / incipient attack time has been called into question. Comments from the API Subcommittee on Corrosion and Materials reports covering the University of Wisconsin work in the 1960‟s and Shewmon‟s work indicate that attack actually starts immediately upon exposure and the incubation period is actually the time it takes to find evidence of attack by whatever detection method is used. If this is the case, the issue is attack propagation. Some plant and laboratory studies have reported that methane has been detected as soon as equipment/samples have been exposed. Although there are some cases in which several hundred thousand hours were required for detectable attack to show up. It has been reported that once attack was detected, it propagated very rapidly. The attack data on C-0.5 Mo from the MPC/API database and the JPI database are plotted against the 1977 curve as follows. 198HFigure A-9 covers the attack points for bare C-0.5 Mo and 199HFigure A-10 shows the experience for clad and weld overlay on C-0.5 Mo backing material. 200HFigure A-11 includes the data sources. Many of the reported incidents with C-0.5 Mo were found after roughly 60,000 hours or longer. Some attack was found after over 200,000 hours. Nelson stated that his original 1949 C-0.5 Mo curve came from data just becoming available from hydrogen processing equipment built during World War II. This would put total time in service in the 50,000 hour range at the most. It should also be noted that inspectors did not have the relatively sensitive inspection tools we have today. Two instances of C-0.5 Mo attack on nozzles show that whether or not the nozzle is corrosion protected with austenitic clad or weld overlay, the thicker the cross section the deeper the attack. This shows that hydrogen concentration is an important attack factor. However, many of the attack problems on C-0.5 Mo seem to have been on piping which is relatively thin wall. Possibly, this is because of piping system stresses. In general, austenitic clad or austenitic weld overlay that is not breached by corrosion or some cracking mechanism, adds hydrogen attack resistance to components made of materials such as C-0.5Mo. This may explain why few attack problems have been found on hydrodesulfurizer reactors and feed/effluent exchangers. Most problems have occurred on unprotected C-0.5 Mo which tended to be used more often in noncorrosive service like Catalytic Reformers. Cladding or weld overlay doesn‟t increase the base material‟s ability to resist hydrogen attack. Point 49 is the Getty point where attack of C-0.5 Mo welds overlaid with Type 309 was found while there was no attack on the base metal which was protected with plug welded 12Cr. The interesting thing about the attack in this and a previously reported Getty incident was associated with transverse weld cracking which probably occurred during fabrication. Two Tosco points (Point 63 and Point 64) were somewhat similar in that attack occurred under a rollbond 12Cr cladding in areas that had considerable blistering. In this case, the operating temperatures covered a considerable range. Attack on a C-0.5 Mo base metal with 347 weld overlay self-reinforced reactor nozzle was found due to blisters noticed in the overlay (Point 65).


144

Another difficulty that could affect fitness for service analysis and the ability of the inspection techniques to find hydrogen attack is the apparent variability of the mode of attack on C-0.5Mo. Some attack problems have been only in weld and/or the heat affected zones while others have been in the base metal away from any known welds, arc strikes or apparent high stress areas. In addition, attack on some thick nozzles has been fairly uniform “layer” type while other incidents have been isolated small areas that the attack has progressed right through the wall with no apparent attack on either side of the small area. Attack has not necessarily been in the highest temperature areas. One of the few carbon steel points for which we have detailed information is available (Point 19) indicates that the attack was very spotty. This was also noted in the detailed information on the C-0.5Mo Point 27 and Point 36. A small area might be attacked part way or even through wall with the material on either side showing no evidence of attack. The spotty nature of high temperature hydrogen attack is very apparent on C-0.5Mo. We have much more information available on C-0.5Mo than any of the other materials. Attack can be spotty and can occur in the weld, heat affected zone or base metal on equipment and piping whether or not it has been post weld heat treated. This apparent variability makes inspection decisions and risk prioritization very difficult and inspection potentially very costly. Obviously, stress affects the attack situation. Inspections should consider the high stressed areas.


145

5.0 CONCLUSIONS

Over fifty years of experience with Nelson‟s curves and API Publication 941 have shown that they have accomplished their original purpose of serving as a design guide for minimizing high temperature hydrogen attack in the refining and petrochemical industry. The inspection procedures used by the industry are constantly evolving and have minimized complete failures that could be blamed solely on high temperature hydrogen attack. The reported attack incidents involve a very small percentage of the equipment in hydrogen service in the refining and petrochemical industry. Even if the most complete data point information is used, it is very difficult to envision using plant data to try and “verify” the curve positions unless a model can be developed from other sources. The use of the available laboratory data would be better for a model or in a verification project since the variables that are inherent in plant data can be controlled.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

146

6.0

FIG

UR

ES

Figu

re A

-1 -

Bou

ndar

y C

ondi

tions

for H

ydro

gen

Atta

ck o

n C

arbo

n St

eel

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

147

Figu

re A

-2 -

Ope

ratin

g Li

mits

for C

arbo

n an

d A

lloy

Stee

ls in

Con

tact

with

Hyd

roge

n at

Hig

h Te

mpe

ratu

re a

nd P

ress

ures

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

148

Figu

re A

-3 -

Ope

ratin

g Li

mits

for C

arbo

n an

d A

lloy

Stee

ls in

Con

tact

with

Hyd

roge

n at

Hig

h Te

mpe

ratu

re a

nd P

ress

ures

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

149

Figu

re A

-4 -

Ope

ratin

g Li

mits

for S

teel

s In

Hyd

roge

n Se

rvic

e

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

150

Figu

re A

-5 -

Ope

ratin

g Li

mits

for S

teel

s In

Hyd

roge

n Se

rvic

e

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

151

Figu

re A

-6 -

Ope

ratin

g Li

mits

for S

teel

s In

Hyd

roge

n Se

rvic

e to

Avo

id D

ecar

buriz

atio

n an

d Fi

ssur

ing

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

152

Figu

re A

-7 -

Expe

rienc

e w

ith C

-0.5

Mo

and

Mn-

0.5M

o St

eels

in H

igh

Tem

pera

ture

Hyd

roge

n Se

rvic

e

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

153

Figu

re A

-8 -

C-0

.5M

o 1

949

thru

199

7

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

154

Figu

re A

-9 -

Hyd

roge

n A

ttack

Exp

erie

nce

with

C-0

.5M

o

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

155

Figu

re A

-10

- Hyd

roge

n A

ttack

Exp

erie

nce

with

C-0

.5M

o - C

lad/

Ove

rlaid


156

Figure A-11 - Hydrogen Attack Experience with C-0.5Mo - Clad/Overlaid - Reference Listing of Data Sources


157

APPENDIX B – COMMONLY ASKED QUESTIONS


158

Additional background items issues are briefly presented here to acquaint the reader with some of the key factors involved in the technical basis.

1.0 HOW DO WE GET CARBON ACTIVITY AND CONTENT?

Carbon activity is not readily measured. The equilibrium relation for methane formation from hydrogen and carbon indicates that methane fugacity is proportional to carbon activity in the ferritic solution, so it is important. A good estimate is all that is needed for most work because the potential methane pressure has a much lower dependence on carbon content or activity than does fugacity. Models in recent decades have avoided discussion of carbon activity. The free energies of the equilibrium mixed carbides in the steel have been used instead as the basis for calculating the fugacity and methane pressure. The activity of carbon in equilibrium with the mixed carbide may be calculated from the free energy change. However, it is simpler to estimate the amount of carbon in solution after heat treatment and assume the concentration is proportional to or approximately equal to activity. To obtain expressions for the methane pressures in steels, the carbon activities must be specified. The carbon solubilities in contact with cementite and graphite in iron are, respectively (25):

[B.1]

[B.2]

The activity of carbon in a plain-carbon steel containing cementite was reported by Stone (Ref. 1)

[B.3]

Carbon activities in N+T and annealed 2-1/4Cr-1Mo steels were measured at temperatures between 900ºF and 1300ºF. (Ref 2) Specimens were equilibrated in liquid sodium containing carbon at various activities. The carbon activities in the steels were reported to be independent of temperature to within the factor-of-two experimental scatter and depended upon carbon concentration in the following manner:

[B.4]

where: C = the weight percent concentration of carbon in 2-1/4Cr-1Mo steel.


159

For this work, this equation is viewed as reasonable and gives values that are not temperature dependent. A plot of activity versus dissolved carbon content , Figure B-1 below, compares Equation B.4 with the simple assumption that carbon activity is equal to the amount of carbon dissolved in the ferrite expressed in weight percent. Within experimental errors or the validity of equilibrium, free energy calculation for mixed carbides after heat treatment the simple estimate that activity is equal the concentration in weight percent is satisfactory for the broad range of low alloy steels of interest in API RP 941 from 1 Cr to 3 Cr contents with molybdenum.


160

FIGURE B-1

ACTIVITY VERSUS DISSOLVED CARBON CONCENTRATION


161

1.1 How Much Carbon Is Necessary To Exhaust The Carbide Formers?

A typical steel contains at least 1 atomic percent carbon since metallic elements have atomic weight 5 to 8 times that of carbon. In the case of C – 0.5% Mo steel, only about .03 wt% carbon is needed to consume all the molybdenum if it has precipitated as Mo2C. The balance of carbon is available to form relatively unstable carbides of the cementite type. Actually, molybdenum is usually dissolved in the other carbides, designated such as M3C and M23C6. Molybdenum‟s effect on activity when dissolved in carbides does not appear to be sufficient to prevent hydrogen attack of some C – 0.5% Mo steels at conditions slightly more severe than for carbon steel. For chrome-moly steels with complex carbides, at least 15 times the weight % of carbon is necessary to precipitate most of the carbon as carbides that are more stable than M3C. This group includes alloys like 1Cr–1/2 Mo steel and 1-1/4Cr–1/2 Mo-silicon steels and extends up to 2-1/4Cr–1Mo steel. For steels with less than 3% total alloy element content, heat treatment, homogeneity and the like may have a significant effect on behavior, unless carbon content is unusually low. For example, a low alloy steel with 0.15 % carbon content will form carbides consuming about 2.5% by weight of the metal. Thus, 1-1/4 Cr–1/2 Mo steel (1.75% total) would be expected to have some iron-rich M3C while 2-1/4 Cr – 1 Mo steel should have no unstable M3C carbides. This would be true unless the material was heat treated at a sufficiently low temperature and remained far from equilibrium.


162

2.0 ARE THE API RP 941 CURVES WHERE THEY BELONG?

It depends on the material and how the curves are to be employed. There is scant evidence that the carbon steel line is not conservative. However, the C – 0.5% Mo lines used in past years were clearly nonconservative. It should be anticipated that for other materials, the curves will not apply under very adverse or unusual conditions of microstructure, composition, heat treatment or applied stress. The possibility of attack is most significant in applications where the actual operating conditions are close to or at the Nelson curve, and where stress relieving, tempering or PWHT have been minimal and any applied, thermal or residual stresses are high. For example, MPC has reported attack of 2-1/4Cr–1Mo steel weldments in only 2,000 psi hydrogen at 825ºF when stress was applied and tempering left the material in a high strength (high carbon activity) condition (about 105 ksi U.T.S.).


163

3.0 WHAT VARIABLES MIGHT ADVERSELY INFLUENCE BEHAVIOR IN LONG TERM SERVICE?

The following table summarizes the likely effect of a number of variables of resistance to hydrogen attack.

Carbon Content – Where carbide stabilizing alloy content is lean, carbon levels at the top of specified ranges might accelerate reaction kinetics, all other things being equal. Also, heat treatment or thermal cycles such as welding that leave the steel supersaturated with dissolved carbon can be expected to accelerate attack.

Large Grain Size – Detrimental effects are reported for coarse grain heat affected zones after welding.

Stress – High stresses have been shown to cause or accelerate attack in the laboratory and the field (e.g. at girth welds in poorly supported piping systems)

Thickness – Thick sections reduce the concentration gradient of hydrogen assuring higher levels of dissolved hydrogen at greater depths in bare and clad or overlayed steels.

Corrosion – May either hinder or promote hydrogen dissolution. However, a surface layer of corrosion products may act as a barrier and retard hydrogen entry.

Large Inclusions or Other Internal Discontinuities –

Weak or low energy interfaces provide free surfaces at which methane may form and enable nucleation of voids or blistering.


164

4.0 WHAT OTHER FACTORS MIGHT PLAY A ROLE IN THE APPEARANCE OF HYDROGEN ATTACK?

Microstructure – Coarse weakly bonded particles such as inclusions are potential nucleation and reaction initiation sites. Coarse pearlite, or other carbides are detrimental.

Creep Strength – Most models recognize that suppression of the steel matrix creep rate by increasing alloy content or decreasing temperature is beneficial.

Hydrogen Content – It is reasonable that reaction rate will depend on hydrogen content. Factors which impede hydrogen entry may result in less than the theoretical maximum amount of dissolved hydrogen (based on Sievert‟s Law). Coatings, surface treatments, poor mass transfer, etc. can impede hydrogen entry into and egress from the steel. Components exposed to hydrogen on two or more surfaces (e.g. tube sheets) are likely to display increased susceptibility.


165

5.0 WHY IS HYDROGEN ATTACK SEEMINGLY UNPREDICTABLE?

There may be significant heat-to-heat differences in behavior even when heat treatments are the same. Some heats react more rapidly than others. Void nucleation rates may differ markedly, depending on impurity content, applied stresses, and heat treatment. Little is understood about the void nucleation process. At low temperatures, the progress of attack is governed by the cavity nucleation rate. Nucleation is a complex function of local stresses, thermal history, surface energy, microstructural characteristics and possibly events involving hydrocarbon intermediaries in the methane reaction. Finally, there are not yet validated, comprehensive, detailed equations to describe nucleation of hydrogen attack.


166

6.0 WHY HAVEN’T THE NECESSARY CRITICAL EXPERIMENTS BEEN RUN?

The potential for long term HTHA of refinery equipment is of practical concern, but long-term experiments in hydrogen are inconvenient, costly and dangerous. Investigators usually do not run long term tests because of these impediments. Instead, they accelerate attack by increasing temperature and/or pressure and using samples exposed to hydrogen on all surfaces for pre-selected, usually short times. The result is a poor simulation of service conditions that typically involve single-side exposures for long times at temperatures where nucleation and void growth rates are governing. Also, data have not been systematically analyzed since critical variables such as, methane pressure, carbon activity, hydrogen content and the like are not easily measured, monitored, understood or controlled. The result is a patchwork of data and models and few, long term systematic studies with adequate information to develop the necessary predictive tools.


167

7.0 WHAT IS ACTUALLY HAPPENING DURING EARLY STATES OF HYDROGEN ATTACK?

Most models now consider void expansion and coalescence under the stress associated with the methane pressure in the voids. In service, cavities are actually quite small. Sometimes cavities much less than a micron in diameter have coalesced to completely cover the fracture surface at failure. They were probably an order of magnitude smaller at initiation of the growth stage. There is little data concerning increases in methane content or cavity volume with time. However, interrupted tests haven‟t been conducted to obtain data on the growth of cavities with time. It has been conjectured that there is an incubation time and Nelson provided plots showing lines interpreted to indicate incubation times as functions of temperatures and pressures. However, for some materials severe degradation has been observed in less time than indicated by Nelson while for others much longer exposure times were required before damage was first detected.


168

8.0 REFERENCES 1. D. Stone Ph. D., Thesis 'Hydrogen Attack Kinetics', Cornell University, 1985. 2. K. Natesan et al., Nuclear Technology, Vol. 28, p. 44,1 March, 1976.


169

APPENDIX C – ESTIMATING DAMAGE RATES FOR LIFE ASSESSMENT

Authors Note: The method described in this section is preliminary. It is included here for purposes of guiding research and discussions and has not been approved by any API committee for use.


170

The main body of the report makes the following key points with regard to life prediction:

The behavioral curve for each alloy in Nelson‟s P/T space is comprised of three main segments. 1) A somewhat horizontal portion attributed to the rate of methane reaction. 2) A substantially vertical portion attributed to creep rate control where reaction rate is rapid. 3) A transition region for conditions lying between 1) and 2)

201HFigure C-1 and 202HFigure C-2 may be helpful in understanding Segment 1. 203HFigure C-1 depicts the behavior at carbon concentrations typical of the common alloys found in API RP 941. The lines are drawn to represent the temperature/pressure conditions for a low, constant damage rate for each of the steels. The materials differ from one another with regard to the carbon concentration in solution in each. Where carbon concentration is low, higher temperatures and pressures (hydrogen contents) are required to achieve the damaging reaction rate. Also, in the computation allowance is made in the activation energy for the free energy change associated with methane formation from either graphite or cementite, as appropriate for each material. Finally, the reaction rates were adjusted on the basis of typical microstructures anticipated. Where carbides are coarse (annealed carbon steel or C-0.5% Mo) many nucleation sites are anticipated and a low rate of reaction gives the target damage rate. Carbides formed in a quenched and tempered material are expected to be fine and offer few nuclei. The same damage rate would require higher hydrogen contents - i.e higher temperatures and pressures to accomplish the same total methane production rate. Essentially, this adjustment accounts for different reaction surface areas in the formulation of Grabke and Martin cited in the report body. The correction ratios were:

0.10% 2 0.08% 3 0.05% 4 0.03% 5

204HFigure C-2 shows how the rate of damage for one (carbon) steel increase with P/T conditions above the limit curve. Lines corresponding to progressive doubling of the reaction rates are shown. If the limit line is chosen to be 2x 10-6/hr then the rates for any other condition may be obtained by interpolation. 205HFigure C-3 and 206HFigure C-4 depict an approach to damage calculation where creep rate controls and reaction rates are rapid enough to cause very high methane pressures . The equilibrium methane pressures are shown on the Y axis for carbon steel (C-3) or carbon 0.5% Mo steel (C-4) at any condition identified by the intersection of curves of constant hydrogen pressure and the (vertical) line for temperatures emanating from the X axis. Intermediate values may be obtained by interpolation between the temperature and hydrogen pressure lines. Overlaid on these coordinates are lines of estimated constant hourly damage rates calculated with the aid of the Omega based strain rate equations found for the respective steels in API 579. Absent other data on damage rate this approach might be used to objectively estimate damage resulting from pressure and/or temperature excursions. This approach can only be justified where externally applied stresses are low and the methane pressure-applied stress interaction discussed in the text can be ignored.


171

Comprehensive information on the effects of methane pressure on the damage rate might be deduced from stress-rupture data for 2 1/4 Cr- 1 Mo steel in hydrogen. Using the proposed interaction equation in this report, it is possible to build a database of predicted failure times for this alloy under the action of methane pressure alone i.e. tHm as shown by the equations below. When stress rupture data are available in both air and hydrogen environments, tH may be calculated from the difference of the reciprocals as shown in Equation C-2.

[C-1]

[C-2]

To demonstrate the approach, calculations were made using the stress rupture data discussed in this report. The results are presented in 207HFigure C-5. Given the variety of sources, heat treatments and test conditions, the scatter is considered reasonable. Essentially, this is a plot of estimated life in hydrogen at the indicated methane pressures (absent significant applied stresses). Methane pressures applicable to other situations may be estimated from the numerous figures in the report. 208HFigure C-6 includes the damage rates suggested by the best fit line of the data in 209HFigure C-5. Locating operating conditions in the region defined by the temperature and constant pressure lines provides an estimate of the hourly damage rate for the exposure. This is the same approach applied in 210HFigure C-3 and 211HFigure C-4. Application of this damage assessment method requires judgments to be made be with regard to microstructure, strength, heat treatment etc as discussed in the report. An activity is now in progress, with API and MPC funding, to systematize an approach for incorporation in RBI and FFS evaluations.


172

FIGURE C-1 – REACTION RATE LIMIT LINES FOR INDICATED CARBON AVAILABILITY. HIGH FIGURES (E.G. 100%) CORRESPOND TO REACTIONS WITH CEMENTITE. DECIMAL NUMBERS

ARE ACTIVITIES OF FOR CARBON IN SOLUTION IN LOW ALLOY STEEL


173

FIGURE C-2 –LINES INDICATING INCREASING RATES OF REACTION RELATIVE TO THE BASELINE FOR CARBON STEEL WITH CEMENTITE


174

FIGURE C-3 – POTENTIAL METHANE PRESSURES FOR CARBON STEEL AT TEMPERATURE AND HYDROGEN PRESSURE COMBINATIONS. LINES OF CONSTANT DAMAGE RATE ARE

ESTIMATED FROM THE MPC OMEGA CREEP RATE EQUATIONS IN API RP 579


175

FIGURE C-4 – POTENTIAL METHANE PRESSURES FOR C-0.5% MOLY STEEL AT TEMPERATURE AND HYDROGEN PRESSURE COMBINATIONS. LINES OF CONSTANT DAMAGE RATE ARE

ESTIMATED FROM THE MPC OMEGA CREEP RATE EQUATIONS IN API RP 579


176

FIGURE C-5 – LARSON-MILLER PARAMETER PLOT OF CALCULATED FAILURE TIMES IN HIGH PRESSURE HYDROGEN FOR 2 1/4 CR 1MO STEEL FREE OF EXTERNAL STRESS. LIFE

ESTIMATES ARE BASED ON PROPOSED INTERACTION EQUATION


177

FIGURE C-6 – POTENTIAL METHANE PRESSURES FOR 2 1/4 CR 1 MO STEEL AT TEMPERATURE

AND HYDROGEN PRESSURE COMBINATIONS. LINES OF CONSTANT DAMAGE RATE ARE ESTIMATED WITH AID OF MEAN LMP LINE IN FIGURE C-5


178

APPENDIX D – EFFECTIVE PRESSURES OF HYDROGEN IN STEEL COVERED BY CLAD/OVERLAY AND/OR

CORROSION PRODUCT


179

Very low diffusivity of hydrogen in stainless clad or overlay materials used for corrosion protection results in an effective pressure of hydrogen at the bond line that is lower that the process stream. This effective pressure is calculated as follows:

For the geometry shown


180

Given:

= Hydrogen flux associated with clad or corrosion covered component,

respectively. = Hydrogen partial pressure in operating environment.

= Hydrogen effective pressure at the clad/overlay surface closest to environment.

= Hydrogen effective pressure at the backing steel interface closest to environment.

= Diffusivities of hydrogen in clad/overlay and backing steel, respectively.

= Thicknesses of clad/overlay and backing steel, respectively.

= Solubility of hydrogen in clad/overlay and backing steel at 1 psi

hydrogen partial pressure, respectively. Terms are dependent on temperature and expressed in appropriate units.

= Pressure reduction ratio at backing steel interface due to presence of clad/overlay - i.e. /

= Pressure reduction ratio at steel interface in the presence of corrosion product. i.e., / in corrosion situations

= Coefficient used to relate flux to pressure differential for corrosion product. Assumed independent of temperature and pressure.

For clad/overlay on steel, the pressure reduction behind the cladding, RC, when there is no hydrogen behind the backing steel we write from Fick‟s First law:

[D.1]

Then, rearranging terms:

[D.2]


181

if settings

[D.3]

gives

[D.4]

and

[D.5]

and

[D.6]

The maximum concentration of hydrogen in the backing steel then is proportional to R 1/2 while

the local methane fugacity decreases in proportion to the decrease in R2. For corrosion products on the clad/overlay surface

[D.7]

by analogy with the system with no corrosion

[D.8]

which yields

[D.9]

or

[D.10]


182

which leads to after rearranging terms

[D.11]

or

[D.12]

The parameter is obtain by monitoring flux under corroding and non-corroding situations. If we define the ratio, J, of the reductions in pressure at the backing steel interface with corrosion (RCP) and without corrosion RC as:

[D.13]

[D.14]

or

[D.15]

Then

[D.16]


183

[D.17]

which gives for

[D.18]

The quantity is obtained by comparing permeation (flux) with and without corrosion product on the clad/overlay surface noting that:

[D.19]

thus:

[D.20]

And in the presence of corrosion covered clad (overlay)

[D.21]

If the corrosion product reduces flux by 20%, then the hydrogen content is reduced by a further 20% as compared to the cladding alone. For example, if cladding itself is expected to reduce the hydrogen content by 25%, then the total effect is a 40% reduction. [1-(1 –.25) (1-.2)]


184

APPENDIX E – OBSTACLES TO UNDERSTANDING HTHA


185

Listed below are some important reasons why advances in understanding HTHA have been slow in coming.

1) Operating records from refineries and chemical plants are not sufficiently detailed or

accurate to be used to establish models or otherwise benchmark the curves. Generally, records maintained have been imprecise indicators of the actual operating history that may contribute to damage of the steels i.e. the hydrogen content, temperature, stress, metallurgical conditions, and even component geometry.

2) The API database only records observed damage - i.e., usually severe attack. Records of

slowly progressing attack would be the most valuable in establishing the threshold times for attack. Components with minor or subcritical damage are likely to go undetected. Therefore, they are not included in API‟s database.

3) Materials of construction have not been well characterized. Composition, heat treatment

and metallurgical history significantly influence behavior, but were seldom documented. 4) Attack is difficult to detect and characterize. Even in advanced stages, HTHA may go

undetected by NDE or even destructive means, including good optical microscopy. Thus, the evolution of damage has not been characterized. Additionally, there is no accepted system for quantifying damage.

5) Susceptibility to HTHA may vary greatly with heat treatment from lot to lot and, possibly,

locally within a lot of material due to stress, metallurgical factors or compositional segregation.

6) The probability of attack is most likely related to the amount of hydrogen dissolved in the

steel during exposure. However, hydrogen content of steel in service is seldom known with certainty. It is rare that any efforts are made to measure it. Even when hydrogen partial pressure and temperature of the process stream are known, the amount of hydrogen actually dissolved in the steel is a function of adsorption efficiency, surface condition, gas-solid interactions, steel thickness and other variables that are difficult to quantify. 0F

* 7) Austenitic cladding retards ingress of hydrogen to the backing steel, thereby lowering the

hydrogen content. This barrier effect may not have been adequately considered by Nelson in plotting some of the original operating data. Such an omission would have resulted in lines that were nonconservative or otherwise misleading.

8) Methane at the pressures of interest is a non-ideal gas and calculating the equilibrium

pressure that may develop based on thermodynamic principles utilizes an equation that must be solved numerically. This has been an impediment to engineers and all early researchers seeking to correlate or even rationalize the driving forces and behavior.

* We will not spend time here considering the wide variety of microstructural hydrogen traps, but steel subjected thermal cycles while in hydrogen service could conceivably build-up high concentrations of high pressure molecular hydrogen at discontinuities within the steel. These could act as methane reaction initiation sites. Traps might be imperfections caused by hot or cold work, inclusions, coarse grain boundary carbides, triple points, etc,


186

9) The pressure of methane that may occur within the steel is important to understanding HTHA. It has been proposed herein that the possibility of forming cavities or cracks is related to the methane pressure in comparison to the local tensile or creep strength of the steel. However, the rate of methane accumulation and actual pressures depend on solid state or gas-solid reactions for which kinetics and thermodynamics are not well understood.

10) Fundamental properties of hydrogen in steel such as diffusivity and solubility vary with

microstructure, composition, heat treatment, and other metallurgical variables. Even when service conditions are known, calculations about reaction rates cannot be made with certainty. Predictions or explanations of behavior must be considered questionable. Explanations of past behavior and failures are often speculative.

11) Perhaps the most important factor in calculating the potential methane pressure, the

thermodynamic activity of carbon, is not readily measured or adequately understood. As a result, there is uncertainty about the calculations of the methane pressure that may actually develop.

12) Owners usually do not monitor hydrogen permeation or content in refinery equipment.

Without quantitative information on the actual exposure conditions, explanations or prediction of behavior are not reliable.

13) Most investigators treat HTHA as a thermodynamic issue. However, nucleation and growth

of voids have important kinetic components which must be understood in order to properly calculate the rate of attack and fitness-for-service of equipment.

14) The time to produce a given amount of methane and the effect of methane on

microstructure and properties are essential to understanding safe operating limits. However, they are seldom measured, even in laboratory studies.


187

APPENDIX F – APPLICATION AND SUMMARY OF PW PARAMETRIC MODEL

PREPARED BY T. NOMURA AND T. SAKAI

(MP) AUTHOR‟S NOTE: THIS SECTION WAS PREPARED BY THE ABOVE NOTED AUTHORS AND REPRESENTS A PROPOSAL MADE AT A POINT IN TIME FOR THE PURPOSE OF DISCUSSION BY THE API COMMITTEE. IT IS INCLUDED HERE FOR THE PURPOSE OF HISTORICAL COMPLETENESS AND IS NOT ENDORSED BY API.


188

Three basic approaches have been tried to provide a technical basis for predicting HTHA:

1) the kinetics of bubble growth, 2) the development of "Arrhenius type" expressions, and 3) a parametric approach using what is called a "Pw" parameter.

The kinetics approach may ultimately provide the most complete understanding of the HTHA process. Unfortunately, the growth kinetics have not yet been established, and this approach is currently not amenable to predicting service lives of our equipment. However, this paper shows that much of the theoretical kinetics work reduces to an Arrhenius form, and that in turn, the Arrhenius form can be usefully characterized by the "Pw" parameter. Alternatively, the "Arrhenius type" expression and the "Pw" parameter provide relatively simple approaches. Shewmon and others [Ref. 2–6] showed experimentally that the strain rate of test specimens accompanied with bubble growth can be described by the Arrhenius type expression. A joint study of the JPVRC (Japan Pressure Vessel Research Council) [Ref. 7] showed that time-dependent critical conditions of 0.5Mo steel can be described by a relatively simple expression named Pw parameter. Later, Nomura and Sakai [8] introduced a modified and more general form of the Pw parameter. Each of these expressions contains only two parameters and the parameters can be determined experimentally. In the present paper, the relationship between the theoretical approaches and the "Pw" parametric treatments will be discussed for the purpose of giving a basis to the Pw parameter. Then, the Pw parameter will be applied to assess the time-dependent limit curves of several steels, the effects of applied stress and the benefits of austenitic stainless steel overlay.


189

1.0 SUMMARY OF THEORETICAL KINETIC TREATMENTS ON BUBBLE GROWTH

Many theoretical treatments have been proposed on the bubble growth kinetics of HTHA. They adopt theories that were developed for cavity growth under externally applied stress or creep cavity growth, which can be categorized into: 1) diffusion model and 2) coupling model between diffusion and power-law creep. This adoption is valid because, in general, externally applied stress is mathematically equivalent to methane pressure in the bubbles. The diffusion model consists of models of grain boundary diffusion, surface diffusion, and coupled grain boundary and surface diffusion. The coupled diffusion and power-law creep model includes the complexities of creep. On the one hand, this model indicates that bubble growth is constrained by the need to creep the surrounding material. On the other hand, this model takes into account the accelerating effect of methane-pressure triaxiality on the bubble growth. The following is a brief outline of these models:

1.1 Diffusion Models - Grain Boundary and Surface Diffusion Early work by Vitovec [Ref. 9] and Shewmon et al. [Ref. 2, 3, 9, 10] adopted grain boundary diffusion models of Hull and Rimmer [Ref. 11], and Raj and Ashby [Ref. 12] that had been developed for creep cavity growth. In this model, the controlling process of bubble growth is the diffusion of iron atoms from the bubble surface to a grain boundary. Here, the diffusion of iron atoms on the bubble surface is assumed to be fast enough for the bubble to keep its equilibrium shape. Sundararajan and Shewmon [Ref. 13] and Parthasarathy [Ref. 14] adopted surface diffusion models by Chuang et al. [Ref. 15]. Nomura and Sakai [Ref. 8] also applied this model to describe the bubble growth of 2.25Cr-1Mo steel, where bubble growth was directly measured by SEM observation.

1.2 Coupled Diffusion and Creep Model A model for constraining bubble growth has been discussed [Ref. 13, 16]. Where bubble distribution is heterogeneous, and there are uncavitated grain boundary facets adjacent to a cavitated facet, the uncavitated facets impose a constraint or back stress on the bubble growth until the creep accommodates the constraint. Odette et al. [Ref. 16] observed the heterogeneous bubble growth in 2.25Cr-1Mo steel and analyzed the growth by adopting an equation of Wilkinson and Ashby [Ref. 17]. The effect of stress triaxiality in the coupled diffusion and creep model has also been examined. Sham and Needleman [Ref. 18] conducted detailed analysis on the effect of stress triaxiality and concluded that the triaxiality accelerates the cavity growth. Van der Giessen et al. [Ref. 19, 20] applied their model to the bubble growth by considering the fact that the triaxiality in methane bubbles is extremely high and even infinite when only methane pressure exists.


190

To summarize, it is evident that these models use the fundamental processes of diffusion (either grain-boundary or surface diffusion), and creep. The equations produced by these models will now be considered more quantitatively.


191

2.0 RELATIONSHIP BETWEEN THEORETICAL EQUATIONS AND ARRHENIUS TYPE EXPRESSIONS

Before reviewing the theoretical equations, it is important to note that Shewmon et al's experimental work on HSLA (High Strength Low Alloy) and carbon steels [Ref. 2–5] and on 2.25Cr-1Mo steel [Ref. 6]. This work showed that the strain rate d /dt due to bubble growth can be described by an Arrhenius type expression:

[F.1]

where: Q = apparent activation energy, PH2 = hydrogen pressure, R = gas constant (1.987 cal/mol K or 8.3145 J/mol K), T = temperature in K, A and = constants The theoretical kinetic equations follow this same form:

As was shown above, several theoretical treatments have been proposed for describing bubble growth. Most of them are based on the processes of diffusion, power-law creep, or coupled diffusion and creep. These basic processes are proportional to two factors, diffusion coefficient and methane pressure with various power exponents as follows:

2.1 Grain Boundary Diffusion (GBD) Process

[F.2]

[F.3]

[F.4]

where: R = bubble radius, Db, Eb = grain boundary diffusion coefficient of iron atom and its activation energy, PCH4 = methane pressure, R = gas constant (8.3145 J/mol K), T = temperature in Kelvin


192

2.2 Surface Diffusion (SD) Process

[F.5]

[F.6]

[F.7]

or

[F.8]

where:

Ds, Es = Surface diffusion coefficient of iron atom and its activation energy.

2.3 Coupling Between GBD and SD Processes

[F.9]

2.4 Power-Law Creep Process

[F.10]

[F.11]

[F.12]

where: Dv, Ev = Lattice diffusion coefficient of iron atom and its activation energy, N = Creep exponent ( = 6.9).

2.5 General Form of All Processes Summarizing all of the equations above, they can be written in a more general form:

[F.13]

[F.14]


193

where:

D = Db or Ds or Dv, and E = Eb or Es or Ev

The relationship between the methane pressure PCH4 (which is the driving force of HTHA) and hydrogen pressure PH2 (which is what we know from service or laboratory conditions) is:

[F.15]

[F.16]

where:

Ko = equilibrium constant for the reaction between graphite and hydrogen to form methane, and ac = the carbon activity in the steel. This relationship is based on the work of Odette and Vagarali [Ref. 22]. For low methane pressures, the ideal gas law is a good approximation of real behavior, and we will show later that for this case the hydrogen pressure exponent m is 2. However, as hydrogen pressure increases, m decreases (212HFigure F-2 and 213HFigure F-4). By using this relationship, all processes lead to:

[F.17]

where:

[F.18]

[F.19]

As a result, bubble growth rate dr/dt can be described by the Arrhenius type expression:

[F.20]

Strain rate due to bubble growth d/dt is proportional to dr/dt [Ref. 14] and thus Equation F.1 is obtained, restated below for convenience:

[F.21]


194

3.0 DERIVATION OF ENGINEERING PARAMETER PW

The "Pw" parameters can now be easily derived. Integrating the general Arrhenius expression

[F.22]

Taking natural logarithm leads:

[F.23]

Defining Pw as follows:

[F.22]

The expression of a general form of the Pw parameter can now be obtained:

[F.25]

It should be noted that, by definition, Pw is proportional to "–ln()", and can be used as a critical value of the damage by bubble growth. It is also important to note that the parameters and Q are the same ones in the Arrhenius type expression.

If it is assumed = 3, Q = 190 kJ/mol, and "log" is used instead of "ln", the Pw parameter as originally used by the JPVRC can be obtained from Equation F.25 (Note that last term becomes Q/2.3RT 9918/T).

[F.23]

The above derivation is based on the Arrhenius type expression. More precisely, the Pw parameter can be derived directly from theoretical equations themselves as follows. In this case, derivation is done from the equation for coupled grain boundary and surface diffusion process, but the parameter Pw can be also derived from the other theoretical equations. The equation by Chuang et al. [Ref. 13,15] can be described as follows:

[F.24]

where: d = a ratio of bubble radius r over half bubble spacing b or (r/b), y = a function of grain boundary- and surface- diffusion coefficients of iron atoms, y = absolute temperature, Ko = the equilibrium constant for the reaction between graphite and hydrogen to form methane, and


195

ac = the carbon activity in the steel.

The temperature dependence of the various factors is as follows:

[F.25]

[F.26]

[F.27]

[F.28]

A = 1/T can be neglected compared with the exponential dependence on temperature, and Equation F.27 is reduced to:

[F.29]

[F.30]

where:

in Equation F.18 = (3/2) in this case (or = (3/2)m),

Q = E – Qm in Equation F.19,and

F' includes factor F which is a function of bubble radius r and the number of bubbles. Therefore, the elapsed time t in which the size of bubbles reaches r is:

[F.31]

where, ro is the initial bubble radius or the radius of pre-existing cavities that are usually assumed.

Including the proportional constant in F" and taking the natural logarithm of Equation F.34 gives:

[F.35]

[F.32]

where:


196

F" = an index of bubble radius or damage of the grain boundary, and Pw = qualified as a parameter which indicates the degree of HTHA. Thus, the Pw parameter can be explained in terms of Chuang et al's theoretical work.

The Pw parameter can also be derived from the other theoretical equations for grain boundary diffusion, surface diffusion, and power-law creep processes with modification of a coefficient of factor "ln(PH2)" and the value of Q. The general form for all the processes is:

[F.33]

4.0 EXAMPLE OF EXPERIMENTAL RESULTS RELATIVE TO THE ARRHENIUS TYPE EXPRESSION

It was shown above that the various controlling processes lead to the Arrhenius type expression, and experimental results or strain rate due to bubble growth can be described by the expression. To show how the expression works, and to show the present state of HTHA study on the carbon steels, experimental examples are given below. Experiments give the parameter (hydrogen pressure exponent) and Q (apparent activation energy) while the required parameters for determining the process are (methane pressure exponent) and E (activation energy of iron atom diffusion). The relationship between these parameters was shown above as Equation F.18 and (Equation F.19 ):

[F.38]

[F.39]

If the values and E are obtained from and Q, it is evident which process among Equations F.2 to F.12 is controlling the bubble growth. The following table summarizing Shewmon's experimental results on Q and , and his calculated results for E and Equation F.39 The steels are electroslag refined A516 steel (No.1, 2) and Al-killed 1020 steel (No.3-5).


197

Table F-1

Experiment Number Temp. (ºC) PH2 (MPa) Q1) (kJ/mol)

1) E2)

(kJ/mol) 2)

1 400 – 500 2 - 5 115 1.9 220 1.5 2 370 – 430 4 - 20 210 0.6 225 1.2 3 400 – 550 2 - 5 112 2.15 220 1.6 4 270 – 375 to 20.8 115 1.0 125 3.3 5 270 – 375 to 20.8 184 3.3 280 9.8

1) Experimental results, 2) Calculated results

Experimental condition for No.1 and No.3 is high temperature (400oC to 550ºC) and low hydrogen pressure (2 – 5 MPa). The condition of No.2 is intermediate temperature (370 to 430ºC) and high hydrogen pressure (around 4 – 20 MPa). Under these conditions, the values of E and agree satisfactorily with those expected for boundary-diffusion-controlled growth. Cases No.4 and No.5 correspond to low temperature (270oC to 375ºC) and high hydrogen pressure (up to 20.8 MPa), but short and long exposure time, respectively. Controlling processes of these cases seem to be surface diffusion and power-low creep, respectively. The resulting strain of No.5 was very large at 310-3. According to Shewmon, the critical damage or "incubation period" corresponds to a strain of about 410-4-10-3. After this much strain, the expansion rate accelerates sharply [Ref. 3] and bubbles link up to form a single bubble over the grain-boundary segment [Ref. 5]. This means that the controlling process during the incubation period is boundary or surface diffusion, and after the incubation period, matrix creep acts as a controlling process. Shewmon [Ref. 5] also noted that the estimated values of E and shown above agree reasonably with those of the model prediction for boundary and surface diffusion processes, but are consistently somewhat higher than the values expected. This could be explained by a small lattice-creep contribution from the regions between the individual bubbles on a bubbled boundary, or from creep accommodation of the unbubbled boundary segments. Shewmon's comment gives an example of assessing the coupling effect between diffusion and power-law creep.


198

5.0 APPLICATION OF THE PW PARAMETER – PREDICTION OF TIME DEPENDENT CRITICAL CURVES

5.1 Discussion of the Values of and Q It was shown above that the generalized Pw parameter can be derived approximately from theoretical equations for various controlling processes, and that the Pw parameter is qualified as a parameter that indicates the degree of bubble growth or damage. In the general form of Pw, = m and Q principally depends on the controlling process and hydrogen condition (pressure and temperature). Therefore, the controlling process and hydrogen condition dependence of Q values. However, in the present study, the same values of 1.5 and 190 kJ/mol for and Q are adopted, respectively, for all steels. Our reasoning is:

3) As was introduced above, a series of experimental and theoretical studies by Shewmon et

al. on carbon steels show that the controlling process is boundary or surface diffusion over the temperature range of 270oC to 550ºC and hydrogen pressure of 2 to 20.8 MPa. Power-law creep process controls only after the "incubation period" (large strains). In the case of 2.25Cr-1Mo steel, it is predicted [Ref. 14] that power-law creep process becomes dominant beyond 20 MPa of hydrogen pressure. Most petroleum industry equipment operate below this pressure. Therefore, we neglect power-law creep processes. When only diffusion processes are considered the value of is 1 for boundary diffusion, 2-3 for surface diffusion, or 1.5 for coupled boundary and surface diffusion. The value 1.5 is near the mean value and it approximately serve for all the diffusion processes.

4) For the value of Q (apparent activation energy), the JPVRC joint study showed that the

value 190 kJ/mol is appropriate in the case of 0.5Mo steel [Ref. 7]. Nomura and Sakai [Ref. 8] also obtained about 190 kJ/mol (45.2 kcal/mol) for the activation energy of bubble growth in 2.25Cr-1Mo steel through the direct SEM observation of bubble growth. This value is also close to the 210 kJ/mol apparent activation energy found for the No.2 carbon steel specimen in Shewmon's work, which was obtained under the conditions of most relevant to operating plants. Furthermore, Sundararajan and Shewmon [Ref. 2] report that all data on three kinds of HSLA (High Strength Low Alloy) steels and one plain carbon steel give the almost the same energy of 188 to 196 kJ/mol, in the temperature range from about 270 to 500ºC at 21 MPa hydrogen pressure.

5) The assumption of 190 kJ/mol apparent activation energy can also be supported by a

theoretical basis. To predict the apparent activation energy Q = E – Qm Equation F.39 or the energy obtained by experiments, the value Qm in the following expression must be evaluated.

[F.40]

[F.41]


199

For carbon steel, the relationship between the hydrogen pressure and methane pressure in equilibrium with carbides can be calculated by assuming that the carbide in the carbon steel is Fe3C. In this case, Koac K is known as follows [Ref. 22,23]:

[F.34]

Adopting the relationship between fugacity and pressure of methane by Odette and Vagarali [Ref. 22], we can calculate the curves in 214HFigure F-1. From 215HFigure F-1, we can evaluate the value Qm that depends on hydrogen pressure and temperature. 216HFigure F-2 shows the hydrogen pressure exponent m calculated from 217HFigure F-1. 218HFigure F-3 shows the result for the case of 2.25Cr-1Mo steel. Calculation was performed by using Equation F.40 with adopting Shewmon's experimental value ac = 0.1 [5] and Ko below.

[F.35]

[F.36]

Ko is the equilibrium constant for the reaction between graphite and hydrogen to form methane and G is the formation energy [Ref. 23].

The relationship between fugacity and pressure of methane by Odette and Vagarali [Ref. 22] is also used in the calculation as is the case of 219HFigure F-1.

220HFigure F-4 shows the hydrogen pressure exponent m in Equation F.40

Now the apparent activation energy Q can be estimated. 221HFigure F-5 shows the results for carbon steel. Here, the cases of two controlling processes, (1) the coupling between GBD and SD processes and (2) the GBD process, are shown.

6) In the case of the coupling process, activation energy is E = 0.5(Eb - Es) + Eb from Equation

F.39, and = 1.5. If we adopt Eb = 192 kJ/mol [Ref. 4] and Es = 130 kJ/mol [Ref. 4], the energy E is 223 kJ/mol and the energy Q is calculated as a function of hydrogen pressure and temperature. Two solid lines in 222HFigure F-5 correspond to the average values over the temperature range of 200ºC to 250ºC and 400ºC to 450ºC. The energy Q lies between 170 to 215 kJ/mol except for the hydrogen condition of high temperatures and low pressures below 2 MPa. This suggests that the assumption of Q = 190 kJ/mol is reasonable for a wide range of conditions. However, if Eb = 206 kJ/mol [Ref. 14] and Es = 130 kJ/mol is adopted, the energy E is 244 kJ/mol and the energy Q is obtained as the dotted lines in 223HFigure F-5. The values are larger by 20 kJ compared to the solid lines and the assumed value of 190 kJ/mol seems to be underestimated.


200

7) Results of the case of GBD process with Eb = 206 kJ/mol [Ref. 14] and = 1.0 are also shown in 224HFigure F-5 as dashed-dotted lines. 190 kJ is reasonable in this case except for the hydrogen condition of low pressures and high temperatures. If Eb = 192 kJ/mol [Ref. 4] and = 1.0 is adopted, the lines are lowered by 14 kJ from the lines for 206 kJ (this case is not plotted in 225HFigure F-5), but the assumption of 190 kJ/mol seems to be still reasonable.

226HFigure F-6 is the result on 2.25Cr-1Mo steel. Definition of solid, dotted, and dashed-dotted lines are the same as the case of 227HFigure F-5. The apparent activation energy ranges from 160 to 210 kJ/mol at 30 MPa of hydrogen pressure and the assumption of 190 kJ/mol seems to be valid, but the value Q gradually decreases with decrease in hydrogen pressure. At 10 MPa, the energy ranges from about 120 to 180 kJ/mol, which is lower than 190 kJ/mol. However, the data for Eb and Es are for pure iron, and it is probable that the assumption of 190 kJ/mol is valid in the case of commercial low alloy steels. The effect of lower value of Q will be further discussed later. In the case of surface diffusion process, very low energy Q is obtained because = 3 is large (the result is not shown in 228HFigure F-5 and 229HFigure F-6). If there is a case where this process is controls, the assumption of 190 kJ will not be appropriate. According to Shewmon's experiment on the carbon steel, which was introduced in the previous section, the surface diffusion controls only at high hydrogen pressure of 20.8 MPa, while, below 20 MPa, grain boundary diffusion controls the process. In the case of 2.25Cr-1Mo steel, observed bubble shape is not crack-like, but usually elliptical [Ref. 8] although the experimental conditions are limited. This suggests that the surface diffusion does not control the process also in 2.25Cr-1Mo steel. As a result of discussion above, the assumption of 190 kJ/mol for apparent activation energy Q will be reasonable with some limitations, which will be further discussed in the following each case.

5.2 Discussion of API RP 941 versus Parametric Equations for Several Materials

5.2.1 Characteristics of Carbon Steel Curves in API RP 941 Time-dependent curves and a portion of the API RP 941 curves [Ref. 1] are reproduced in 230HFigure F-7 as dotted lines. Those curves have the following characteristic:

8) At higher hydrogen pressures, the curves are weakly dependent on hydrogen pressure and

show nearly horizontal lines. 9) At lower hydrogen pressures, the curves are strongly dependent on hydrogen pressure and

show nearly vertical lines.

It should be noted that the curves at 100 h and 200 h are drawn primarily on Weiner's experimental data [Ref. 24], noted as "E".

5.2.2 Using the Pw Parameter to Generate Carbon Steel Curves As was explained above, calculation was conducted by using the Pw parameter that is defined as follows:


201

[F.45]

where: PH2: MPa, t: h, T: K, = 1.5m (Note this is the same as Equation F.39, if = m is used. As discussed earlier in this section, = 1.5 and Q = 190 kJ/mol are reasonable values for carbon steel). The critical Pw was determined so that the 100000 h line crosses the "No attack line" of API RP 941 at 1000 psi and 500ºF, i.e., Pw = 25.65 was adopted. As for the methane pressure exponent m, the average value at temperatures from 250oC to 350ºC and hydrogen pressures of 5 to 20 MPa was adopted. 231HFigure F-1 gives the value of about 0.25, or = 0.375. These values give the following expression:

[F.46]

The results for various times are shown as dotted lines in 232HFigure F-7. The curves are nearly horizontal and the 100 h-curve is close to the API 100 h curve at higher hydrogen pressures. However, the Pw curves differences as compared to the API curves. One difference is the interval between the lines or time dependence. In the case of the Pw lines, the interval increases as the time increases, while the API curves become close together with times, and the lines for 1000 and 10000 h lie close to the no-attack line. This difference will be discussed in the next section for 0.5Mo steel. Another difference is the significant deviation at lower hydrogen pressures. Part of the difference at low pressures is because for simplicity we assumed that m & Q were constant. However, 233HFigure F-2 and 234HFigure F-3 show that m and Q take larger and smaller values respectively at low hydrogen pressures. However, the change in m and Q of 235HFigure F-2 and 236HFigure F-3 does not satisfactorily explain the nearly vertical curves in API RP 941. This feature will be discussed more below.

To obtain better curves for the low pressure region, the following parameters that were experimentally determined by Weiner himself must be adopted [Ref. 24].

5.2.2.1 Weiner's Experiment

Weiner [Ref. 24] analyzed his experimental data on a carbon steel and found that the incubation time t can be expressed as:

[F.47]

where, t: h, PH2: psi, and T: K After converting units, this gives the same equation as Pw with the critical value as follows:

[F.48]


202

where: PH2: MPa, t: h, and T: K

The Pw gives curves that are satisfactory as to Weiner's data that are limited to lower pressures and shorter times, while Pw cannot be applied to the higher pressure region.

237HFigure F-8 shows the coupling curves between the expressions for higher and lower hydrogen pressure region, respectively.

In 238HFigure F-8, the parameter with the values m = 3 and Q = 61 kJ/mol is applied until 8 MPa (1150 psi) of hydrogen pressure for the 100 h-curve, while 239HFigure F-2 and 240HFigure F-5 show that these values are expected only at hydrogen pressure lower than 1 MPa. The inconsistency is a problem to be solved in the future. Another experiment that gives information for the carbon steel will now be described.

5.2.2.2 McKimpson and Shewmon's Experiment

McKimpson and Shewmon [Ref. 3] performed experiments on a carbon steel (Electroslag refined A516) by measuring the expansion rate due to bubble growth that reported that the hydrogen pressure dependence of the rate is significantly different between the regions of low- and high-hydrogen pressure. These correspond to Specimen No.1 and Specimen No.2 in 241HTable F-1:

At low hydrogen pressures [F.49]

At high hydrogen pressures [F.50]

They calculated that the critical damage or "incubation period" corresponds to a strain of about 4x10–4 – 10–3. At greater strains, the expansion rate accelerated sharply and bubbles linked up to form a single bubble over the grain-boundary segment [Ref. 6]. Assuming a conservative value of 4x10–4 for the incipient attack, we calculated time-dependent critical curves from the data above. 242HFigure F-9 shows the results, with the following characteristics: 10) At higher hydrogen pressures, the slope is nearly horizontal. This comes from the small

value of = 0.62, although it is a little larger than the 0.375 used in Equation F.46. The apparent activation energy Q = 210 kJ is also near the 190 kJ used in Equation F.26 These are consistent with 243HFigure F-2 and 244HFigure F-5, although the lines are shifted to a higher temperature, probably due to metallurgical differences.

11) At lower hydrogen pressures, the slope is not as steep as Weiner's case. This reflects the smaller value of = 1.9 that is theoretically reasonable ( 245HFigure F-2), and, in addition,

Q is still smaller but more reasonable (246HFigure F-5) compared with Weiner's case. This suggests the possibility that the steep lines in API RP 941 might be corrected to less steep lines in the future.


203

5.2.3 0.5Mo Steel The joint study of the JPVRC showed that the conventional Pw ( = 3, Q = 190 kJ/mol) gives satisfactory curves, at least for hydrogen pressures of about 700 to 2000 psi [Ref. 7]. 247HFigure F-10 shows the result as compared with the curves of Figure A-3 of API RP 941 [Ref. 1]. Three data from the JPVRC report are plotted, which indicate that the curve for 1000 h should be located close to the curve of JPVRC, and more than 100ºF above the API curve. This suggests a probability that the time dependence of Pw is more accurate than API RP 941 curves. In this case, = 3 (hydrogen pressure exponent m = 2) was adopted. More precisely, the exponent m should be less than 2. In spite of this situation, the slope of the API curves are steeper than that of the JPVRC curves, which is a problem for future investigation.

5.2.4 2.25Cr-1Mo Steel

5.2.4.1 Effect of Bubble Surface Tension

As is shown in 248HFigure F-3, the methane pressure is very low compared with carbon steel, and a sintering force due to surface tension must be considered [Ref. 6].

[F.51]

(Driving force for bubble growth is ) [F.52]

where: Surface tension = 2J/m2 Average bubble radius r = 0.25 microns

The average hydrogen pressure corresponding to the methane pressure 16 MPa is about 3 MPa between 450ºC to 600C and the term "ln(PH2 – 3)" was used in Pw instead of "ln (PH2)".

5.2.4.2 Incubation Curves

249HFigure F-4 shows the hydrogen pressure exponent m. At around 450C to 600C (842F to 1112F), an average m value of about 1.3 ( = 2) at hydrogen pressure of 10 MPa is reasonable. This values = 2 and Q = 190 kJ/mol as described before were used in the calculation. The result is shown in 250HFigure F-11 where the critical value 14.88 was determined by experimental data obtained by Nomura and Sakai [Ref. 8], which are plotted in the figure. This 2.25Cr specimen may have relatively poor resistance to HTHA because of a high Sn content [Ref. 8].

The parameter is as follows:

[F.53]


204

where:

PH2: MPa, t: h, T: K

The dotted lines at temperatures above 1300F mean that the mechanism of damage of steels is not certain at very high temperatures. For instance, surface decarburization and microstructural changes should be considered.

In the calculation above, constant values of m and Q are adopted over the whole range of hydrogen pressure. Actually, these values gradually change with hydrogen pressure as is shown in 251HFigure F-4 and 252HFigure F-6. In order to estimate the effect of the change, three sets of the values were adopted in calculation. They are:

10.37 = –2.8 ln(PH2 –3) – ln(t) + 170,000/(8.3145T) for PH2 < 10MPa [F.54]

13.58 = –2.0 ln(PH2 –3) – ln(t) + 180,000/(8.3145T) for 10 PH2 20MPa [F.55]

17.50 = –1.2 ln(PH2 –3) – ln(t) + 190,000/(8.3145T) for PH2 > 20MPa [F.56]

Here, the critical values were determined so that the curve for 100,000 h is connected smoothly at 10 and 20 MPa of hydrogen pressure. 253HFigure F-12 shows the result. The effect is almost negligible at hydrogen pressure higher than about 1500 psi (10.34 MPa) and Equation F.54 is generally applicable. If detailed information at low-pressure region is required, expression with different m and Q might be necessary, as in the case of carbon steel.


205

6.0 APPLICATION OF THE PW PARAMETER – ESTIMATION OF THE EFFECT OF APPLIED STRESS

6.1 Derivation of Pw Including the Stress

The following is an attempt to estimate the effect of applied stress.

In most theoretical equations, the methane pressure and applied stress is mathematically equivalent, and the bubble growth under both the methane pressure and the stress can be written, where PCH4 is simply replaced with PCH4 + in Equation F.27 More precisely, the proportional factor is not constant, and the factor is a function of bubble radius r or F(r) . Then

[F.57]

where the temperature dependence of the factors is:

[F.58]

[F.59]

[F.60]

[F.61]

Therefore:

[F.62]

From Equation F.62, the following equation can be obtained:

[F.63]

By taking the natural logarithm and putting Pw –ln(F"), = m:

[F.64]

It can be confirmed that EquationF.64 reduces to Equation F.37 in the case of no applied stress ( = 0):

[F.65]


206

6.2 Effect of Applied Stress on the Critical Curve of 2.25Cr-1Mo Steel By using the above Equation F.64, the effect of the applied stress on the critical curve of 2.25Cr-1Mo steel in 254HFigure F-11 can be estimated. As was shown in 255HFigure F-3, the methane pressure of 2.25Cr-1Mo steel is much less than that of carbon steel in 256HFigure F-1, and it can be expected that the applied stress affects the critical curve. The same values that were used in the calculation of 257HFigure F-11 is adopted here - i.e., Pw = 14.88, = 2.0 ( = 1.5), Q = 190 kJ/mol, surface tension = 3 MPa, and the methane pressure of 258HFigure F-3.

The operating or designs condition of reactors is generally: Hydrogen pressure PH2 = 10 – 20MPa Temperature = about 454C (Limit of 2.25Cr-1Mo steel) The methane pressure at the above condition is (259HFigure F-3):

[F.66]

On the other hand, the allowable design stress at 454ºC (ASME Section VIII Div 2) is:

[F.67]

From Equation F.67 and Equation F.68

[F.68]

Therefore:

[F.69]

As a result, it can be seen that the applied stress of 150 MPa has apparently the same effect of increasing hydrogen pressure by about 5 MPa. This means that the critical curve under the effect of applied stress 150 MPa can be obtained for t = 100000 h as

(F.70)

where:

PH2: MPa, t: h, T: K,

The calculated results are shown in 260HFigure F-13. The critical temperature decreases by about 50°F (at PH2 = 10 MPa) to about 25°F (at 20 MPa).


207

7.0 APPLICATION OF THE PW PARAMETER – ESTIMATION OF EFFECT OF OVERLAY

Using the Pw parameter, the effect of overlay on the critical conditions can be easily estimated. The following is an example of the estimation in the case of 2.25Cr-1Mo base metal with an overlay of austenitic stainless steel. Austenitic stainless steel overlays decrease the hydrogen contents of the base metal. It is necessary to calculate the hydrogen contents of the base metal at the boundary between the overlay and base metal, which can be done as shown in the next section.

7.1 Hydrogen Distribution in a Vessel Wall with Overlay During Operation The hydrogen distribution in austenitic () overlay and ferritic () base metal at steady state during operation can be calculated by using conditions at the boundary between overlay and base metal [Ref. 25]:

[F.71]

Hydrogen flux : In(left) = Out(right))

[F.72]

(Local equilibrium)

where: Ce

, Ce: Hydrogen solubility of austenitic overlay and base metal, respectively

D, D: Hydrogen diffusivity of austenitic overlay and base metal, respectively

In the case of stainless steel (Types 309, 347), and 2.25Cr-1Mo steel, these values are as follows [Ref. 26]:

[F.73]

[F.74]

[F.75]

[F.76]

Cb, Cb

=Hydrogen contents of overlay and base metal, respectively, at the boundary between overlay and base metal.


208

P = Hydrogen pressure in the pressure vessel in kgf/cm2

Lo, LB = Thickness of overlay and base metal.

From Equation F.72:

[F.77]

From Equation F.71 and Equation F.77 :

[F.78]

If it is assumed that the hydrogen contents are Ce and zero at the overlay surface and outer

surface of base metal, respectively, the hydrogen distribution in the wall can be calculated. Examples of the results are shown in 261HFigure F-15, where the operating condition is 850°F and 2133 psi (14.7 MPa) for temperature and hydrogen pressure, respectively. The base metal thickness is 15 to 30 cm. When there is no overlay, the hydrogen content at the surface of the base metal is 4.6 ppm, while the overlay reduces the content at the boundary to 3.6 – 3.0 ppm depending on the thickness of the base metal. This means that in evaluating the critical condition, the hydrogen pressure corresponding to hydrogen contents of 3.63 – 3.0 ppm should be used instead of 2133 psi giving 4.6 ppm.

7.2 The Effect of Overlay on the Critical Curve As shown in the previous section, the hydrogen content of the base metal below the overlay Cb

is reduced by the overlay.

[F.79]

The hydrogen pressure PH2 that is in equilibrium with Cb

is lower than the process hydrogen pressure P, and can be obtained from the following relation:

[F.80]

From Equation F.79 and Equation F.80 :

[F.81]

This means that, in the case of pressure vessels with overlay, the critical hydrogen pressure should be determined by PH2 and the critical process hydrogen pressure P is higher than the pressure PH2. The critical process hydrogen pressure P is estimated by putting the PH2 (Relation (8.6) into the equation of Pw with the critical values of Pw of Equation F.70 and process temperature:


209

[F.82]

[F.83]

[F.84]

262HFigure F-14 shows the critical curve as compared with the curves of no-overlay. The expected shift of about 50ºF can be seen. The API RP 941 curve is horizontal (hydrogen-pressure independent) beyond 2000 psi. If this is true, the benefit of the overlay fade away beyond a certain pressure, higher than 2000 psi. 263HFigure F-16 and 264HFigure F-17 present the application of this approach to 2 ¼ Cr ½ Mo steel showing time dependence of behavior and the effects of stress with and without cladding.


210

8.0 CONCLUSIONS

The relationship between theoretical treatments and parametric treatments on HTHA to the give a basis for the Pw parameter has been discussed. Then, the Pw parameter was used to assess the time-dependent limit curves of several steels, the effect of applied stress and finally the benefits of austenitic stainless steel overlay. Summarizing :

1) The general form of the Pw parameter is consistent with theoretical treatments and also

with an experimentally derived Arrhenius type expression. This clarified the theoretical and physical meaning of the hydrogen pressure exponent , and the apparent activation energy Q that constitute the Pw parameter. These factors principally depend on which controlling process of methane-bubble growth is active, carbon activity (or steel species) and hydrogen conditions.

2) The values and Q were estimated for carbon and 2.25Cr-1Mo steels by calculating the

methane pressure and applying the relationship of and Q with methane pressure exponent and activation energy E, which identify the controlling process.

3) Based on the results above, time-dependent critical curves were assessed for carbon and

2.25Cr-1Mo steels. The results for 0.5Mo steel by earlier JPVRC work were also shown. In this evaluation, different values of were adopted, while the same value of 190 kJ/mol was applied except for the low hydrogen-pressure region of carbon steel. At low hydrogen pressures of carbon steel, extremely low values of what cannot be theoretically predicted must be used if we try to follow the near vertical curves of API RP 941.

4) The effect of applied stress was estimated for 2.25Cr-1Mo steel. The stress of 150 MPa that

corresponds to the allowable stress at 454°C (850°F), lowers the critical temperature by about 50 to 25°F at 10 to 20 MPa of hydrogen pressure.

5) The effect of austenitic overlay was estimated for 2.25Cr-1Mo steel. The overlay raises the

critical temperature by about 50°F.

In the present assessment of time-dependent critical curves, effects of metallurgical factors were not discussed. It is expected that several metallurgical factors affect the critical conditions. The assessment of these effects remains for future work.


211

9.0 REFERENCES

1) Steels for Hydrogen Service at Elevated Temperatures and Pressures in Petroleum Refineries and Petrochemical Plants, API Recommended Practice 941, Fifth Edition, January 1997, Supplement 1, April 1998, American Petroleum Institute.

2) G.Sundararajan and P.G.Shewmon: Metall. Trans.A, 1980, vol.11A, p. 509 3) M.McKimpson and P.G.Shewmon, Metall.Trans., 1981, Vol.12A, p.825 4) B.Panda and P.G.Shewmon, Metall. Trans., 1984, Vol.15A, p.487 5) P.G.Shewmon, Materials Science and Technology, 1985, Vol.1, p.2 6) T.A.Parthasarathy, H.F.Lopez and P.G.Shewmon, Metall.Trans., 1985, Vol.16A, p.1143 7) Hydrogen Attack Limit of C-0.5Mo Steel, JPVRC, May 1987 8) T.Nomura and T.Sakai, "Theoretical and Parametric Methods for Evaluating Hydrogen

Attack Limit", ASME PVP-Vol. 359, 1997, p.315 9) F.H. Vitovec: Am. Petrol. Inst. Proc., 44, (3), (1964), p.179 10) P.G.Shewmon: Metall. Trans.A, 1976, vol.7A, p. 279 11) D. Hull and D. E. Rimmer: Phil. Mag. 4. 673 (1959). 12) R. Raj and M. F. Ashby: Acta Metall, 23, (1975), 653 13) G.Sundararajan and P.G.Shewmon: Metall. Trans.A, 1981, vol.12A, p. 1761 14) T. A. Parthasarathy: Acta metall. 33, 1673 (1985). 15) T.-J. Chuang. K. I. Kagawa. J. R. Rice and L. B.Sills: Acta Metall., 30, (1979), p.265 16) G. R. Odette. B. L. Chao and G.E. Lucas: "Kinetics and Mechanism of Hydrogen Attack in

2.25Cr-1Mo Steel” -- Part I-III, ORNL/Sub/82-22276/01, August 1988 17) Wilkinson and Ashby: Acta Metall., 23, (1975), p. 1277 18) T.-L. Sham and A. Needleman: Acta metall. 31, (1983), p.919 19) E. van der Giessen, M.W.D van der Burg, A. Needleman and V. Tvergaard; J.Mech. Phys.

Solids, 43, (1995), p.123 20) M. W. D. van der Burg, E. van der Giessen and R. C. Brouwer: Acta Metall., 44, (1996),

p.505 21) L.Martinez and W.D. Nix: Metall. Trans. A vol. 13A, (1982), p.427 22) G.R. Odette and S.S. Vagarali: Metall. Trans A, 13A, (1982), p. 299 23) G. H. Geiger and O. F. Angeles: "Study of Effects of High-Temperature, High-Pressure

Hydrogen on Low-Alloy Steels," API Publ. 945, (1975) 24) L.C. Weiner, "Kinetics and Mechanism of Hydrogen Attack of Steel," Corrosion, 1961,

Vol.17, p.109 25) T.Sakai et al, "Hydrogen Distribution through Pressure Vessel Wall," JPVRC Doc.

No.MHE-6, Presented at PVRC Meeting in Scottsdale, Jan. 26, 1981 26) T.Sakai et al, "Effect of Hydrogen on MPT and De-hydrogenation during Shut Down in

Hydroprocessing Reactors," ASME PVP-Vol.1, 1997, p.79


212

10.0 FIGURES

Figure F-1 Hydrogen Pressure Vs. Methane Pressure for Carbon Activity for Fe3C


213

Figure F-2 Hydrogen Pressure Exponent of m of Methane Pressure for Fe3C


214

Figure F-3 – Hydrogen Pressure vs. Methane Pressure for Carbon Activity ac=0.1


215

Figure F-4 Hydrogen Pressure Exponent m of Methane Pressure for ac=0.1


216

Figure F-5 Apparent Activation Energy Q for Carbon Steel


217

Figure F-6 Apparent Activation Energy Q for 2.25Cr-1Mo Steel


218

Figure F-7 Time for Incipient Attack of Carbon Steel with Hydrogen exponent m+0.25


219

Figure F-8 Time for Incipient Attack of Carbon Steel by Coupled Equations


220

Figure F-9 Time for Incipient Attack of Carbon Steel by Shewmon. Incipient Attack Means Link-up of Bubbles at Strain of 0.0004.


221

Figure F-10 Time for Incipient Attack of 0.5Mo Steels


222

Figure F-11 Time for Incipient Attack of 2.25Cr-1Mo Steel


223

Figure F-12 Time for Incipient Attack of 2.25Cr-1Mo Steel. Different m and Q for Three Regions of Hydrogen Pressure are Adopted.


224


225

Figure F-13 Effect of Applied Stress on the Critical Curve of 2.25Cr-1mo Steel


226

Figure F-14 Effect of Austenitic Overlay on the Critical Curve of 2.25Cr-1mo Steel


227

Figure F15 – Hydrogen Distribution During Operation with the Effect of Austenitic Overly


228

Figure F-16 2-1/4Cr-1Mo Steel Hydrogen Damage Lines Based on Equation F-53


229

Figure F-17 Summary 100,000 Hours Damage Lines for a 2-1/4Cr-1Mo Steel Vessel


230

Figure References 1. M.W.D. van der Burg, E. van der Giessen and V. Tvergaard, “A Continuum Damage Analysis of

Hydrogen Attack in a 2-1/4Cr-1Mo Vessel.” 2. American Petroleum Institute Publication 941 3. M.W.D. van der Burg, “Hydrogen Attack; Micromechanical Modeling on Three Length Scales”, Thesis

Delft University 1997. 4. Tadamichi Sakai, Tohru Nomura and E.H. Niccols, “The Basis and Application of the Pw Parameter

for High Temperature Hydrogen Attack”, 2000 ASME Pressure Vessel and Piping Conference, July 23-27, 2000, The Westin Hotel, Seattle, Washington.

5. T. Sakai et al, “Effect of Hydrogen on MPT and Dehydrogenating During Shut Down in Hydroprocessing Reactors,” ASME PVP, Vol. 1, 1997, p.79.


231

APPENDIX G – OVERVIEW OF EUROPEAN RESEARCH

(MP) AUTHOR‟S NOTE: THIS SECTION CONTAINS INFORMATION DEVELOPED BY MANY INVESTIGATORS AND SUBMITTED TO API OVER A PERIOD OF TIME FOR THE PURPOSE OF DISCUSSION AND EVALUATION BY THE API COMMITTEE. IT IS INCLUDED HERE FOR THE PURPOSE OF HISTORICAL COMPLETENESS AND IS NOT ENDORSED BY API.


232

1.0 INTRODUCTION

This section is intended to acquaint the reader with the advanced modeling concepts developed over a period of years especially by the team at Delft University of Technology. There are three sections. The first contains abstracts showing the range of topics being explored. The second presents the content of a paper of M.W.D. van der burg and E. van der Giessen published in Materials Science and Engineering [A220 pp 200-214, 1996)] It elegantly sets forth most of the important concepts required for a model that includes interaction of diffusive void growth and creep. It is an excellent starting point for anyone interested in pursuing an analytical approach to HTHA. Finally, the third part is a thoughtful discussion prepared for this project by Delft‟s S.M. Schlogl and E. Van der Giessen. It systematically explores and contrasts the work reported in Appendix G with that of the investigators at Delft. It enables the reader to grasp the points of agreement and the sensitivity of conclusions to assumptions made. Below are presented abstracts of a number of the papers showing the range of detail studied at Delft. Further details on the references are found at the end of this Appendix.

1.1 "General Introduction to the Micromechanical Analysis of Hydrogen Attack on Three Length Scales"

Abstract – In the (petro-)chemical industry, the operating conditions of components functioning under high temperature and high hydrogen pressure are determined by hydrogen attack (HA). This material degradation process is investigated for 2-1/4Cr-1Mo steel, an industrial standard material for pressure vessels. The material degradation is due to the reaction of dissolved hydrogen in the steel with grain boundary carbides. The subsequently formed methane cannot diffuse so cavities develop at the carbide locations. When the grain boundary cavities coalesce, a microcrack forms preluding intergranular rupture. The cavitation process is modeled on three length scales, where results of a lower length scale serve as input for the higher length scales. Starting at the length scales of a single cavity, the evolution of the representative cavity is analyzed with detailed Finite Element (FE) calculations. With a cavity growth relation capturing the numerical growth rates, HAA on polycrystal size level can be investigated with FE analyses. These polycrystal results were captured by a continuum damage relation which serves as the constitutive relation for FE vessel analyses, which is the third, specimen size length scale. Since initial assumptions in the modeling affect the lifetime predictions, the most important are mentioned explicitly here.

1.2 E. van der Giessen, M.W.D. van der Burg, A. Needleman, and V. Tvergaard, "Void Growth Due to Creep and Grain Boundary Diffusion at High Triaxialities," 1994. Abstract – The growth of grain boundary voids at elevated temperatures by coupled creep and grain boundary diffusion is studied numerically using a cylindrical unit cell model. Emphasis is on the influence of the remote stress triaxiality, which is taken to cover the full range of axisymmetric stress states, from purely effective to purely hydrostatic states of stress. The motivation for extending previous results stems from the need for an accurate cavity growth model to analyze damage due to hydrogen attack, where the grain boundary voids are internally pressurized. Because of the wide range of stress states considered, numerical stability requires the use of two normalizations of the variation principle for the coupled void growth problem: one when the effective stress is dominant and the other when the mean stress is dominant. In the regime where deformation is primarily by creep, two distinct modes of deformation appear for


233

each level of porosity: one for low triaxialities and one that takes over for sufficiently high triaxialities. Approximate models found in the literature for a dilute concentration of voids, or for finite concentrations, are explored to check their ability to represent the stress state dependence of the volumetric void growth rate. A novel approximate formula is derived for creep dominated growth and is shown to give good agreement with numerically computed void growth rates in the high triaxiality regime and for finite concentrations. A fairly abrupt transition between creep dominated void growth and diffusion dominated void growth is found when the stress triaxiality is very high, so that the interaction between creep and diffusion is then relatively unimportant. Finally, formulae are presented which give an approximate, yet fairly accurate, expression for the void volume growth rate due to coupled diffusional and creep growth over the full range of axisymmetric stress states.

1.3 S. M. Schlögl, Y. van Leeuwen, and E. van der Giessen, "On Methane Generation and Decarburization in Low Alloy Cr-Mo Steels During Hydrogen Attack," Met. Trans. A. Abstract – Low carbon, low alloy Cr-Mo steels may fail by hydrogen attack when they are exposed to high hydrogen pressures at elevated temperatures. During this process, the dissolved hydrogen reacts with the carbides of the steel to form methane in grain boundary cavities. The methane pressure inside these cavities depends on the microstructure of the used steel which consists of a ferritic matrix and alloy carbides such as M7C3, M23C6, M6C, and M2C. The different phases in the multicomponent system Fe-Cr-Mo-V-C are modeled with the sublattice model. Their Gibbs energies are then used to calculate the equilibrium methane pressure in dependence of the microstructure. Driven by the methane pressure, the cavities grow due to grain boundary diffusion and dislocation creep which is described by analytical relations. This leads to progressive development of damage inside the material, but at the same time to a decrease of the carbon content in the steel. This reduction depends, among other factors, on the methane pressure and the damage static. As the carbon content also affects the creep parameters, this process of decarburization may accelerate the cavity growth. Model calculations are used to obtain insight in the influence of this decarburization process on damage evolution and the final lifetime.

1.4 R. C. Brouwer, "Hydrogen Concentration Distributions in the Wall of Pressure Vessels Made of Conventional and V-Modified Steels," The International Journal of Pressure Vessels and Piping, 1992.

Hydrogen diffusion through the wall of clad reactors has been modeled both during normal operation and during shutdown using a finite difference model. The resulting hydrogen concentration distributions have been studied for reactor vessels constructed of three different steels, one conventional 2-1/4Cr-1Mo steel, and two V-modified steels, 3Cr-1MoV, and 9Cr-1MoV, and the probability of hydrogen assisted crack growth and cladding disbonding has been assessed. The simulation results clearly indicate that under the same operating conditions and using the same cool down procedures the v-modified steels have higher hydrogen concentrations but at the same time a reduced probability for hydrogen assisted crack growth and cladding disbonding. An important implication is that the API limit of 23 motH/m3 for 2-1/4Cr-1Mo steels may not be applicable for V-modified steels.

1.5 M. W. D. van der Burg, and E. van der Giessen, "A Continuum Damage Relation for Hydrogen Attack Cavitation," Acta mater., Vol. 45, No. 7, pp 3047-3057, 1997.


234

Abstract – A Continuum Damage Relation (CDR) is proposed to describe the failure process of hydrogen attack, i.e. grain boundary cavitation of steels under conditions of high temperature and high hydrogen pressure. The cavitation is caused by the chemical reaction of hydrogen with grain boundary carbides forming cavities filled with high pressure methane. The micromechanisms described are the grain boundary cavitation and the dislocation creep of the grains. The CDR is based on two extreme cavitation rate distribution modes. In the first mode, the cavitation rate along the facets is uniform, resulting in a hydrostatic dilatation while the creep deformations remain relatively small. In the second mode, cavitation proceeds predominantly on grain boundary facets transverse to the principal macroscopic stress. This part of the CDR builds on Tvergaard's constitutive relation for intergranular creep rupture [Tvergaard, V., Acta metallurgica, 1984, 32, 1997] where the facet cavitation is constrained by creep of the surrounding grains. The mode corresponding to the highest cavitation rate is the active mode. The two-dimensional version of the CDR is verified against detailed finite element analyses of hydrogen attack in planar polycrystalline aggregates. Finally, the generalization to a three-dimensional CDR is discussed.


235

2.0 NON-UNIFORM HYDROGEN ATTACK CAVITATION AND THE ROLE OF INTERACTION WITH CREEP, M.W.D. VAN DER BURG AND E. VAN DER GIESSEN, DELFT UNIVERSITY OF TECHNOLOGY, LABORATORY FOR ENGINEERING MECHANICS, DELFT,(THE NETHERLANDS)

Abstract – Hydrogen Attack (HA) is the development of grain boundary porosity by cavities filled with high pressure methane that originates from the reaction of carbides with hydrogen at high temperatures. The cavities grow by grain boundary diffusion and by creep of the adjacent grain material till they coalesce with neighboring cavities to form a microcrack. Earlier work on HA has focused on unit cells containing a single cavity, using average cavitation properties. Here, non-uniform cavitation properties on the grain size scale are assumed in a polycrystalline aggregate, and unit cell analyses are performed to investigate the influence of the adjacent grains on the development of the grain boundary HA. The numerical results are explained in terms of two simplified models which highlight the key parameters governing the grain deformation–grain boundary cavitation interaction process.

2.1 Introduction Hydrogen attack (HA) is a well-known low-ductility type of failure in steels exposed to high pressure hydrogen at elevated temperatures as is relevant in, for example, petrochemical applications [1]. A failure due to HA can be recognized by intergranular fracture, which is initiated by porous damage of the grain boundary facet. This damage develops as a result of the fact that at these elevated temperatures, hydrogen can diffuse into the steel, where a chemical reaction takes place with the carbides. Due to this reaction, a cavity is formed that is filled by methane gas, which is trapped in the material. Depending on the stability of the carbide, the equilibrium methane pressure can be of the order of the ambient hydrogen pressure, or two orders of magnitude higher (see e.g. [2]). Together with any applied stresses from the loading of the component, the internal pressure drives growth of the cavities. Because of the occurrence of grain boundary diffusion, the cavities located at the grain boundary facets develop faster than cavities inside a grain. Also dislocation creep can contribute to the growth of the cavities. The grain boundary cavitation proceeds until the cavities coalesce to form a microcrack, and linking up of the microcracks leads to intergranular macroscopic failure.

Important earlier investigations of HA focused on the evolution of one single cavity. Cylindrical or spherical unit cells containing one central cavity were investigated using analytical relations for the diffusive cavity growth, in some cases also accounting for creep deformations. In the analyses of Shih and Johnson [3] and Parthasarathy [4], only the internal pressure was taken into account. Later Shewmon [5] included applied stress (perpendicular to the grain boundary facet) in the model, which tends to accelerate cavity growth, but in this analysis it was assumed that the cavity grows solely by grain boundary diffusion. Building on earlier work [6], [7], Van der Giessen et al. [8] presented a (partially) new cavity growth relation where diffusion and creep are coupled. The relation has been verified with extensive detailed numerical cell model analyses for all possible stress states which may be encountered under HA circumstances. In this way, the combined effect of the internal pressure and the remote applied stresses on cavity growth is incorporated. The relation has been applied in [2] to predict HA failure in 2.25Cr–1Mo steels.


236

These single cavity studies have given first indications on how HA develops in time. However, they imply that the cavities are distributed uniformly along the grain boundary facets. In reality, carbides are not uniformly distributed over a grain facet and, more importantly, the carbides can have different compositions leading to vast differences in the methane pressure of the gas that they may form (e.g. [2], [5]). It is as yet not understood well what the influence is on HA of non-uniform distributions of microstructural properties, like e.g. differences in reactivity of the carbide along a facet. In view of the possible variations in methane pressure distribution along a grain facet, compatibility of deformations is likely to lead to stress redistributions, similar to those associated to Dyson‟s [9] creep constrained cavitation in creep rupture situations. As demonstrated in [9] and later confirmed in many other studies (e.g. [10], [11]), grain boundary cavitation in such cases is slowed down by the surrounding grains, which creep deform very slowly. Thus, the results of full field simulations of cavitation in a polycrystalline aggregate [10], [11] were found to differ dramatically from the results obtained from a single cavity model under the applied state of stress.

In order to investigate the effect on HA of microstructural variations as mentioned above, we make the scale transition from the level of individual cavities to the polycrystal level. For that purpose, the two–dimensional polycrystal model of Van der Giessen and Tvergaard [11] is extended for HA. Some preliminary explorations with this HA polycrystal model have been reported briefly in [12] and [13]. The objective of the paper is to draw up a complete picture of HA on the polycrystal scale and to provide an understanding of the influence of carbide variations under different relevant conditions. The polycrystal model itself and the route to the numerical solutions are described in Section 2 and Section 3. The model is then used to study non-uniform HA cavitation on two different size scales. First, in Section 4, internal cavity pressure variations are considered along grain boundary facets, and next, in Section 5, internal cavity pressure variations over the polycrystal aggregate are studied. Two highly simplifying descriptions are discussed subsequently in Section 6 to try to better understand the results of the numerical simulations. These will be shown to capture some key aspects of the interactions that take place in the process of cavitation.

2.2 Polycrystal Model for HA At elevated temperatures in a hydrogen rich environment, the hydrogen molecules dissociate into hydrogen atoms which diffuse into the material. While diffusing through the material, some hydrogen atoms will react with carbides present in the steel. During this reaction, methane molecules are formed which, contrary to the hydrogen atoms, cannot diffuse away through the metal. The (equilibrium) methane gas pressure in this reaction varies substantially with the stability of the carbide, and typically a material contains stable as well as unstable carbides. The equilibrium pressure for stable carbides is of the order of the applied stresses, whereas it can attain values that are orders of magnitudes larger than the applied stress for unstable carbides [14], [15], [2]. The ultimate internal pressure of the methane–hydrogen gas mixture, pm, acting on the cavity surface consists of the partial methane pressure and the partial hydrogen pressure [2].

[G.1]


237

Due to this internal pressure a cavity can develop. The most important cavities are located on the grain boundary facets where they can develop rapidly by grain boundary diffusion, assisted by dislocation creep. Since these grain boundary carbides cause the final intergranular failure, these are the only carbides taken into consideration. The grain boundary cavities grow until they coalesce and form a microcrack, and linking up of the microcracks leads to the macroscopic intergranular crack.

In this paper, it is assumed that hydrogen only reacts with the grain boundary carbides that are present from the beginning. An implication of this assumption is that in our model no new cavities nucleate during HA. This appears to apply well for 2.25Cr–1Mo steels [16]. Another assumption made here is that the methane pressure is constant in time. However, Shih and Johnson [3] have reported that the equilibrium methane pressure will not always be reached in some cases at all moments. Hence, our considerations are expected to overestimate the actual cavity pressure and hence the cavity growth rate. The polycrystal model of Van der Giessen and Tvergaard [11] is used here to simulate the HA process on grain size scale. This model was initially developed to investigate creep rupture by grain boundary cavitation, but since that process is closely related to HA the model can be easily adapted to study HA evolution. In the sequel, we shall only briefly describe the model, with an emphasis on those aspects that are new for the HA application; for details, we refer to [11]. The polycrystal model is based on a two–dimensional polycrystalline aggregate comprising hexagonal grains, as shown in 265HFigure G-1. All grains are assumed here to have the regular hexagonal shape in the undeformed state with facet length 2RI. Periodicity in the microstructure of the polycrystal is exploited by constructing the aggregate on the basis of a unit cell. Assuming also a double microstructural symmetry inside the unit cell (see 266HFigure G-1), only a quarter of the cell has to be analyzed. In the x1- and the x2-direction of the global coordinate system, the polycrystal aggregate is subjected in general to macroscopic applied stresses 1 and 2, as shown in 267HFigure G-1, under plane strain conditions. The grains deform elastically as well as by Norton power law creep in an isotropic manner. The constitutive equations are formulated in a finite strain formulation, using convected coordinates with metric coefficients and gij and Gij in the undeformed and deformed configurations, respectively. The covariant components of the Lagrangian straitens or are denoted by ij, and the conjugate stresses are the contravariant components i j of the Kirchhoff stress tensor on the current base vectors. The total strain-rate is taken to be the sum of the elastic part and

the creep part . Thus, with the elastic stress-strain relationship , in terms of the

Jaumann stress-rate, the constitutive relations for the grain

material can be written as:

[G.2]


238

with:

[G.3]

where e is the effective Mises stress, , and the stress deviator components

are defined by with the mean stress defined by . The effective

creep strain-rate , according to the Norton power law, is given:

[G.4]

Here, 0 is a reference stress parameter, is a reference strain-rate parameter, and n is the creep exponent.

A spherical cap-shaped cavity along a facet can be characterized by its radius a and its equilibrium tip angle (see 268HFigure G-2). The separation between two neighboring cavities is 2b.

The volume of a cavity is , where the function h() is the geometrical cavity

shape parameter defined as:

. [G.5]

At temperatures typical for HA, cavities grow by grain boundary diffusion and by creep of the adjacent grain material. Thus, the volumetric cavity growth rate consists of a part due to grain boundary diffusion, , and a part due to creep, . Growth is driven by the gas pressure pm inside the cavity, but can be accelerated or slowed down by stresses remote from the cavity, as characterized by , and , being the average stress normal to the grain boundary facet, the average Mises effective stress and the average mean stress remote from the cavity, respectively. To determine the volumetric growth rate, a single cavity has been analyzed numerically, first by Needleman and Rice [6] for remote uniaxial stress states, and by Sham and Needleman [7] for stress triaxialities witnessed at macroscopic cracks or notches. These analyses have been extended by Van der Giessen et al. [8] to cover all possible axisymmetric stress states. This elaboration was necessary because in the case of HA, the cavity may grow solely due to the internal cavity pressure, which can be classified as cavity growth under a purely hydrostatic stress state. The numerical results could be captured fairly well by approximate analytical relations, which have been presented in their most complete form in [2]. Only a summary of the relations will be given; more details can be found in the above references.


239

When the cavity grows by creep, two different creep growth modes can be distinguished [8]: a mode which can be associated with low stress triaxialities and/or low porosities (i.e. small values of a/b) and a second mode which can be associated with high stress triaxialities and/or high porosities. The volumetric growth rate by creep for the first mode, according to [2] and [8] is given by:

[G.6]

[G.7]

where the shorthand notation is introduced, and where

. Here, TS is the surface tension and e is equal to the remote

effective stress . The constants n and n are defined by n = 3/(2n) and n = (n-1)(n+0.4319)/n2. Finally, sign(m) denotes the sign of m. In case of the second creep growth mode, for high porosities or stress triaxialities m/e, the volumetric growth rate is given by:

[G.8]

[G.9]


240

The expression for the volumetric growth rate for growth by grain boundary diffusion is a modification of the Hull and Rimmer [17] result for rigid grains, improved for the coupling with creep in the grains [6], [7]. With the extension for the internal cavity pressure pm given in [8], the diffusive growth rate is given by:

[G.10]

where the grain boundary diffusion parameter = DBB/k depends on the boundary diffusivity DBB, on the atomic volume and the energy per atom measure k. The diffusive growth is driven by the average stress normal to the grain boundary facet in addition to the internal cavity pressure pm. The free surface energy S and the surface tension TS commonly result in a sintering effect, expressed through a sintering stress S = (1-f)2S sin/a + f2TS sin/a . However, as soon as the cavity can open and develop, the sintering stress S has hardly any effect on the further evolution [2]. Therefore, both the free surface energy S and the surface tension TS are neglected in the analysis by puttingS = 0. The parameter f in (9) is determined by the diffusive path length Ldiff and the cavity radius a by

[G.11]

When creep in the grains is negligible, the diffusive path length Ldiff is equal to the cavity half-spacing b. However when the creep deformations in the vicinity of the cavity become of the same order of magnitude as the diffusional deformations along the facet [6], then the diffusive path length shortens, so that the volumetric growth rate increases. The type of interaction between creep deformations and the diffusional process depends on the creep modes (indicated with superscript H and L); Van der Giessen et al. [8] have shown that:

[G.12]

is a fair approximation. Here, is a stress and temperature

dependent length parameter, first introduced by Needleman and Rice [6]. The ultimate volumetric growth rate is obtained from (5)–(9) as:


241

[G.13]

The surface diffusion is assumed to be rapid enough to maintain the spherical-caps shape of the cavity. This shape assumption seems to be consistent with the observations for 2.25Cr–1Mo steel by Shewmon [16]. Then the growth rate of the cavity radius a is given by:

[G.14]

with according to (5)–(10).

Due to the growth of the cavities, the adjacent grains separate. If all cavities have the same size and the same spacing, this average separation is determined by:

[G.15]

In a schematic way, 269HFigure G-3 depicts a variation in the separation along the grain boundary facet as a result of differences in e.g. cavity volume. Differentiating with respect to time, gives the average separation rate

[G.16]

Thus the separation rate is a function of the cavity growth rate and of creep deformations in the plane of the facet, expressed in the second term in the right-hand side. The average separation rate will vary along the facet as a consequence of variations caused by variations of cavity distributions and by variations in the volumetric growth rate. The latter are due mainly to variations of the internal driving pressure pm due to different carbide compositions [2], by variations in cavity size, and by a non-uniform remote stress distributions along the facet. Creep in polycrystalline metals may be accompanied by grain boundary sliding [18]. However, in materials that have failed by HA, grain boundary sliding has not been observed. This may be due to the temperature levels at which HA susceptible materials are applied and also to the presence of a high density of carbides on the grain boundaries which tend to act as obstacles for sliding. Therefore, grain boundary sliding is not incorporated into the model. In the model, a micro crack will form when cavities coalesce at a/b = 1. However in reality, grain boundary failure by cleavage of the ligament is expected to occur earlier, this in analogy with creep ruptures by cavitation [18]. It is assumed here that failure takes place at a/b = 0.7. However the precise value of a/b is not very important for the ultimate time to failure, as cavity growth accelerates strongly at such large values of a/b.


242

The results obtained with the polycrystal model will be compared with a single cavity model as discussed in detail in [2]. This model boils down to directly subjecting the representative cavity to the macroscopic stresses 1 and 2. The cavity growth relations in the single cavity model are the same as presented above, but with replacing

, and

by

and ,

respectively. The difference between cavitation in the single cavity model and in the polycrystal model is that in the latter the increase in cavity volume has to be accommodated by deformations of the surrounding grains. The creep resistance of the grains may constrain the volume cavity growth rate and, therefore, the cavitation evolution. This concept of creep constrained cavity growth was pointed out by Dyson [9] for cavitation under circumstances leading to intergranular creep rupture, but is expected to apply equally well here. In the single cavity model, these kind of interactions with adjacent grains are not present, and therefore in the single cavity model, cavity grow this unconstrained.

2.3 Method of Analysis The numerical method used to solve the problem discussed above is largely similar that in [11] and [19]. In this section we therefore only give a brief summary, and refer to [19] for more details. The grains themselves are discretized by quadrilateral finite elements, each one of which consists of four triangular constant strain triangular elements in a „crossed triangle‟ configuration. The finite element grid used is depicted in 270HFigure G-4. The HA and resulting grain boundary cavitation process is incorporated through grain boundary elements. In this case, the only deformation mode for these interface-type elements is the grain separation . Consistent with the finite element representation inside the grains, the grain boundary elements use linear interpolations for and for the other grain boundary characteristics, such as cavity size a and spacing b. For computational convenience, a layer of linear elastic springs is added to the grain boundary elements with a normal stiffness kn, so that the local normal stress n at the grain boundary is determined through:

[G.17]

Here, is the actual separation and is the separation due to cavitation, as being governed by the evolution relation (11). A large value of the stiffness kn ensures that the deviation

remains small; here, we have used kn = 10 E/R, where E is the Young‟s modulus.


243

The governing equations for the grains as well as for the grain boundaries are solved in a linear incremental manner. The remote normal stress to be used in the cavity growth relations (5)

to (9) is taken from the grain boundary stress, i.e. . The remote effective and mean

stress, and , are evaluated as averages from the corresponding stresses in the grain elements adjacent to the grain boundary. To increase the numerical stability of the procedure, a forward gradient approach is used to integrate the constitutive relations (2) and (12). Finite strains are accounted for in the analyses, but in the present applications the strains remain below a few percent. The symmetries imposed on the quarter cell imply that the (quarter) cell boundaries remain straight throughout the process, and shear stress free. Therefore, uniform displacement rates are prescribed in the x1 and x2 directions so that the average true stresses 1 and 2, respectively, retain specified constant values in time.

2.3.1 Variations Along Grain Boundaries In this section, the influence of non-uniformities in internal cavity pressure on the scale of grain boundaries will be investigated. All grain boundary facets in the polycrystal have the same uniform carbide distribution, but the reactivity of the carbides and therefore their resulting internal cavity pressure pm are taken to vary along the facet. The uniform cavity distribution along a facet is characterized by the initial cavity spacing relative to the facet length as (b/R)I = 0.1. The initial cavity size is taken to be (a/b)I = 0.01. In all cases, the Poisson‟s ratio = 0.33, the cavity tip angle = 75º, and the creep exponent n = 5. To describe the variation of the internal cavity pressure over the facet, a local coordinate is introduced along the grain boundary, ranging between 0 and 1. The first internal cavity pressure distribution we are considering is characterized by with , and is shown graphically in 271HFigure G-5. It is assumed that there are no macroscopic stresses, 1 = 2 = 0, so that HA is solely driven by the internal cavity pressure. Note that because of symmetry, cavitation damage will develop equally on each grain boundary facet. The various parameters for creep and diffusion are expressed through the dimensionless group bI/Lm, where Lm is defined by:

[G.18]

In practice [2], the value of this parameter can range from log (bI/Lm) = -2, where cavitation is completely dominated by diffusion, up to around log (bI/Lm) = -2, where the creep contribution to cavity growth is most important. Within this range, three parameter values are taken: (i) log (bI/Lm) = -0.5 where cavity growth is diffusion dominated; (ii), log (bI/Lm) = 0.5 where diffusive cavity growth is expected to interact with creep, and finally (iii) log (bI/Lm) = 1.5 where cavitation is creep dominated. To follow the cavity evolution in time, a reference time scale is introduced by , which is based on the diffusive cavity growth process.


244

An obvious important parameter to look at is grain boundary damage in terms of a/b. For the three cases, the damage evolution is shown in 272HFigure G-6. In 273HFigure G-6 (a) through (c), the a/b distribution is depicted when a/b reaches the value of at the center of the facets (i.e. at about half the ultimate damage). It can be seen that the damage distribution tends to become more non-uniform for higher bI/Lm, i.e. when the creep rate (relative to the diffusional rate) increases. In 274HFigure G-6 (d) through (f), the damage distribution is depicted when first coalescence of cavities occurred (a/b = 0.7) and it is seen that the three distributions differ significantly. In 275HFigure G-6 (d), the cavity-size distribution is virtually uniform, whereas this uniformity has faded in 276HFigure G-6 (e), and is strongly non-uniform in 277HFigure G-6 (f). To compare these polycrystal results with the single cavity model predictions, the damage evolution in the middle and at the triple point of a facet are plotted in 278HFigure G-6 (g) through (i). The single cavity model prediction is shown for an internal cavity pressure of (which corresponds with the maximum internal cavity pressure along a facet), as well as for an internal pressure which corresponds to the average internal pressure . Notice that the differences in damage evolution under these different internal cavity pressure increase strongly with increasing relative importance of the creep growth rate. In all three cases, unconstrained growth of the single cavity under the peak cavity pressure is faster than growth at the facet center in the polycrystal, even in 279HFigure G-6 (i) where the creep rate inside the grains is relatively large. The single cavity model results when the cavity is subjected to the average internal pressure are always in between the polycrystal results at facet center and triple point.

In order to further elucidate the origin of the phenomena observed above, it seems appropriate to also investigate the grain separation () or the separation rate along the facet. The reason is that it is the difference in separation rate between a cavity and its neighboring cavities that must be accommodated by deformations of the adjacent grains. When the resistance of the grains to such deformations is large, stress redistributions inside the grains will take place, leading to „reactive‟ stresses acting on the grain boundary facet, such that compatibility is assured. Even though the present setting for HA is somewhat different, this phenomenon of internal stress redistributions is similar to that underlying Dyson‟s [9] concept of creep constraints on creep rupture. Therefore, we also investigate the evolution of the facet normal stress distribution n() over the facets. For the first case where log (bI/Lm) = -0.5, it can be seen in 280HFigure G-7(a) that the average grain separation is nearly uniform at the moment of first coalescence. This indicates that all cavities have grown at nearly the same rate. This is possible only when the cavity growth rate due to the internal pressure is partially counteracted in some regions or accelerated by normal stresses in other regions, in such a way that the grain separation rate becomes uniform along the facets. In the middle of the grain boundary, a compressive stress constrains the evolution, whereas close to triple points growth is accelerated by tensile normal stress, as can be seen in 281HFigure G-7(b). In this case, where diffusion is much faster than creep, there is insufficient time for the grain to creep deform significantly to accommodate the non-uniform separation rates caused by the cavity pressure only. As a consequence, as shown in 282HFigure G-8 (a), the average separation rates near the triple point and in the middle of the grain boundary are virtually identical, except for an initial transient. 283HFigure G-8 (b) shows the development in time of tensile stresses near the triple point, and of the compressive stresses in the middle of the facet. After an elastic transient, the normal stresses stabilize in time, and the grains separate as virtually rigid bodies.


245

In the next case, for log (bI/Lm) = 0.5, the creep rate is increased relative to the diffusional rate. Note that the separation distribution in 284HFigure G-9 (a) is much more nonuniform than in the case for log (bI/Lm) = -0.5 in 285HFigure G-7(a). Now, there has been sufficient time for the grains to deform by creep flow, thus accommodating the variations in separation rate. At the same time, the magnitude of the normal stresses is seen to be somewhat lower (see 286HFigure G-9 (b)). Comparing the separation rates throughout the process as shown in 287HFigure G-10(a), with the previous results in 288HFigure G-8 (a), we see that the separation rates in the center and at the triple point of a facet now differ significantly. The evolution of the normal stresses shown in 289HFigure G-10 (b) indicate that stress re-distributions take place throughout the process, but the stress levels remain below those in 290HFigure G-8 (b). The normal stress is still of importance in this case since cavity growth is still mainly diffusion controlled. When log (bI/Lm) = 1.5, this is no longer the case. Diffusion is hardly of importance anymore and the process is fully creep dominated. This leads to relatively large creep deformations of the grains in the neighborhood of the highest cavity pressure, as shown clearly in 291HFigure G-11 demonstrating the strongly non-uniform separation distribution along the grain boundary facet. In this creep dominated case, facet normal stresses have lost their importance for cavity growth, but now the distributions of mean stress and effective stress near the grain boundary are of primary importance, as can be concluded from the growth relations (5)–(8). The magnitude of these stresses is much smaller than the stresses built up in the previous cases, but due to the highly nonlinear creep relations these small stresses still result in a significant effect on time to first coalescence compared to fully unconstrained growth, as we have seen in 292HFigure G-6 (i). The previous results have shown that, depending primarily on bI/Lm, non-uniform cavity pressure distributions over grain facets can lead to substantial normal stresses over the grain facets, even though there is no externally applied stress. In these cases the normal stress distribution has to be self-equilibrated, and leads to compressive stresses in the neighborhood of high cavity pressures and tensile stresses elsewhere. This has been seen to lead to constrained growth of the cavities that have relatively high internal gas pressures, and accelerated growth of cavities under low internal pressure. Thus, facet regions of high gas pressure seem to „interact‟ with regions of low pressure through continuous redistributions of normal stress over the facets. An apparent measure of this interaction is the maximum value of

occurring anywhere along the facet: the larger this ratio, the higher the interaction and the stronger the tendency for uniform separation along the facet. By changing the wave length of the methane distribution, it is investigated if the magnitude of this interaction is dependent on the length scale of the pressure distribution. 293HFigure G-12 (a) depicts a distribution of the internal cavity pressure with half the period of the variations, as given by . The previous calculations have been repeated

for this distribution for the intermediate value log (bI/Lm) = 0.5. Comparing the separation distribution at the time of first coalescence, shown in 294HFigure G-12 (b), with 295HFigure G-9 (a), it is observed that the shorter wave length distribution leads to somewhat more uniform cavitation. This suggests that interaction increases with decreasing wave length in the methane distribution.


246

The restriction of using grain boundary elements in the polycrystal model is that only gradients in grain boundary properties along a facet can be accounted for in the analyses. It is not possible to model a single aggressive carbide surrounded by relatively harmless carbides, which may happen in practice. However, especially for these cases, where the grain material creeps severely, one can rely on the single cavity relation expressed in (5) and (6). This is possible because these relations govern the volumetric growth rate of a cavity in an infinite creeping medium. The cavity spacing b is irrelevant in those cases.

2.3.2 Variations Over the Polycrystalline Aggregate In this section, we consider non-uniformities in the internal cavity pressure distribution over the polycrystalline aggregate. In contrast with the previous section, the internal pressure is now constant along any facet, but can differ from facet to facet. Only a single representative microstructure is considered, shown in 296HFigure G-13, as this highlights the issues involved. More complex microstructures have been considered in [12], [13]. The central facet in the unit cell is taken to contain relatively aggressive carbides, characterized by pm/E = 1.0x10-3, whereas on the other facets the carbides are assumed to be harmless; the carbide density and the initial cavity sizes are taken to be the same as on the central facet, but their internal pressure pm is taken to be negligible, i.e. pm = 0 for convenience. In this section, the parameter bI/Lm is taken to vary from log (bI/Lm) = -0.5 to log (bI/Lm) = 0.5. All other parameter values are kept the same as above, including the assumption that no macroscopic stresses are applied, 1 = 2 =0. In 297HFigure G-14 (a) through (c), the distribution of the damage parameter a/b over all facets is plotted just prior to first cavity coalescence. In 298HFigure G-14(a) where log (bI/Lm) = -0.5, the a/b distribution has developed to become virtually uniform. In 299HFigure G-14(b) where the creep rate is enhanced relative to the diffusional rate, log (bI/Lm) = 0.0, the a/b distribution has become slightly non-uniform, while in 300HFigure G-14(c), where the creep rate is even faster, log (bI/Lm) = 0.5, the central facet is nearly completely cavitated whereas on the inclined facet nearly no damage has developed. Note that the parallel top facet in the quarter unit cell, which is further away from the central cavitating facet than the inclined facet is, has cavitated more. Hence, the interaction (in the sense of the previous section) with the top facet has been stronger than with the adjacent inclined facet. In 301HFigure G-14 (d) through (f), the evolution of a/b at the midpoint of each facet is shown. These results seem to give additional evidence for the fact that the interaction with the neighboring facets decreases with increasing value of bI/Lm. Notice that in the case of vanishingly small interaction, cavitation on the facets with the non-aggressive carbides would not develop at all.


247

The interactions referred to here are again most clearly demonstrated by the distributions of the separation and of the facet normal stress n. As shown in 302HFigure G-15 (a), the separation distribution for log (bI/Lm) = -0.5 is almost perfectly uniform, indicating rigid grain behavior. With increasing bI/Lm, 303HFigure G-15 (b) through (c), the distribution becomes more and more non-uniform over the unit cell by significant creep deformation of the grains. It is emphasized that since the cavity pressures on the inclined and the parallel top facet are zero, the only way these facets can cavitate is by virtue of stress redistributions in the grains during evolution of the cavitation on the central facet. In 304HFigure G-15 (d) through (f) the normal stress distribution along the facets are plotted, again just prior to coalescence. In all three cases a compressive normal stress is found on the central facet in order to constrain cavity growth there. In 305HFigure G-15 (d), for the diffusion dominated case, the normal stress on the other facets is tensile. The magnitudes of these internal stresses are such that the separation rate is the same for all facets. In 306HFigure G-15(f), for the larger value bI/Lm, the normal stress distribution is strongly non-uniform in accordance with the non-uniform cavitation. In 307HFigure G-16 (a) through (c), the development of these normal stresses in time are plotted for the center of the facets, and in 308HFigure G-16 (d) through (f) the corresponding evolution of the separation rates normalized by (cf. (13)). The compressive stress counteracting the cavity growth on the central facet is clearly seen now to be maximal for the smaller value of bI/Lm. Also with increasing bI/Lm, the normal stresses keep evolving during the process; thus, the interactions evolve continually during the process. Near the end of the lifetime for the case in 309HFigure G-16 (a), the separation rates become very large, so that grain deformations are negligible; this demands that facets separate at the same rate, as is clearly seen in 310HFigure G-16 (d). In the case where the relative creep rate is highest, log (bI/Lm) = -0.5, it can be seen in 311HFigure G-16 (f) that the separation rates at the inclined and top facet are very small compared to that on the central facet.

The above results indicate that if is large, the creep deformations of the grain are negligible so the grain will behave virtually rigid. A consequence of this is that the separation rates of the different facets are kinematically coupled, as illustrated in 312HFigure G-17. In this figure, the separation rate of facet i, , is coupled with the separation rate of the adjacent facet j, ,

according to /2 = sin . In most cases where the grains behave rigid, cavitation is dominated by diffusion and therefore controlled by the facet normal stress n. Without exploring this further here, these conditions suggest a simple method of analyzing HA cavitation in large polycrystalline aggregates. Assuming uniform carbide distributions and internal pressures along each facet, but at different magnitudes (like in this section), the current (uniform) normal stress over all facets can be determined by satisfying the kinematic couplings of the separation rates. In cases where the rigid grain behavior is present from the beginning, the time to first coalescence can be readily solved for. It seems that the kinematic coupling in case of rigid grain behavior also holds for three-dimensional polycrystal microstructures.


248

2.4 Discussion The results presented herein suggest an intricate connection between the influence of variations in the cavity pressure distribution (i.e. carbide distribution) over a polycrystal and the creep resistance of the material relative to the diffusion rates. To further pin down the essential elements of this, we consider some simplifying descriptions. The first one is a one–dimensional model consisting of two parallel bars. The idea behind this model is similar to Dyson‟s [21] two–bar model that he used to demonstrate creep constrained cavity growth during creep rupture. The two bars are imagined to represent two different parts of the material; one containing aggressive carbides and the other with less aggressive, more stable carbides, as illustrated in 313HFigure G-18(a). The length of the bars, which must be of the order of the grain size, is taken to be 2RI, and in the middle of each bar we imagine a grain boundary facet where cavitation takes place. The main difference between bars A and B is the methane pressure inside the cavities (but also the cavity spacing or the cavity radius may differ). Both bars deform by power law creep and the two parallel bar set may be subjected to an applied uniaxial stress 2 (see 314HFigure G-18 (b)). Due to the difference in the internal cavity pressure, cavitation at both bars will develop with different separation rates and , respectively. In this parallel arrangement however, both bars must elongate at the same rate. This compatibility condition reads:

[G.19]

assuming that A,B << RI. Equation (14) can be directly rewritten as:

[G.20]

From this condition, it is clear that the separation rates must be equal if the grain material does not deform significantly by creep, i.e. . The separation rate

can only become equal by way of stress redistribution among the two bars, so as to constrain the cavity growth rate in bar A and accelerating cavitation in bar B. If the grain (bar) material deformation becomes significant, the separation rates and will differ and the magnitude of the (either accelerating or constraining) internal normal stress is lower. If the grain deformations are not of any significance, the normal stress redistribution will be highest and interaction is maximal. It is of importance at this point to appreciate the role of any applied stress 2. Raising the stress level will affect both the creep in the grain as well as the grain boundary cavitation. However the creep process is accelerated more by the applied stress 2 than the diffusional process because of the strong nonlinearity in the creep relations. So if cavitation is diffusion dominated and the grain (bar) material can creep deform, applying any additional stress 2 will tend to decrease the interaction.


249

If grain deformations are significant, the magnitude of the interaction during the lifetime is not constant. The reason for this is that the separation rate increases very strongly in the course of the cavity evolution (see e.g. in 315HFigure G-8 (a)). As the interaction is changing continuously, there is no single simple parameter to characterize the magnitude of interaction for the whole lifetime, except in the limiting case mentioned above when grain deformations can be neglected all together. Although the two–bar model is illuminating in relation to the phenomena observed in e.g. 316HFigure G-6 and 317HFigure G-13, it cannot explain the influence of gradients of the cavity pressure distribution on the time to first coalescence, shown in the case of 318HFigure G-12. To explain this phenomenon, we consider a second model. When there is a gradient in the separation rate

along the grain boundary facet, due to variations in internal cavity pressure, as illustrated in 319HFigure G-19 (a), this gradient.

[G.21]

has to be accommodated by shearing of the adjacent grain material, as can be seen in 320HFigure G-19 (b). The creep resistance of the grain material will induce a normal stress distribution on the facet, as shown schematically in 321HFigure G-19 (c). These stresses will also tend to constrain rapid cavitation (i.e. large ) and accelerate slow growth (i.e. small ). Also here, if the grain cannot deform significantly, severe stress redistributions have to take place to make the

and uniform. With this simple model, it can be understood how a larger gradient in

cavity pressure distribution gives rise to a larger gradient in which will result in higher interaction stresses that constrain cavity growth from aggressive carbides more, giving a more uniform cavitation. A rough indication about whether or not cavitation will be uniform can be obtained by comparing the separation rate with the creep deformability of the grain expressed in . This leads to

the dimensionless number , where l is the distance over which interaction is investigated. For example, the l can be taken as the wave length of the pressure distribution or as the facet width RI, while can be identified as defined in (G.22). Then, as we have seen in the previous Sections (e.g. 322HFigure G-8, 323HFigure G-10, and 324HFigure G-16), interaction is strong (cavitation is uniform) when log roughly larger than 1, whereas interaction is negligible

when log < -1 or so. When the wave length of the pressure distributions decreases,

increases and therefore the interaction increases, as we have found. When subjected

to applied stresses 1 and 2 (see 325HFigure G-1), we take as a rough measure of the creep deformability of the grain.

[G.22]


250

incorporating creep due to both internal cavity pressure and external stress. For the separation rate only the contribution of diffusion is taken into account (in situations where creep controls cavity growth, the creep rate in the grains is so high that dominates anyway). Then with (G.14) and neglecting the creep term in (G.20), the separation rate is found as:

[G.23]

where d/(f) = ln(l/f) – (3-f) (1-f)/2. Only if f <<1, then d(f) is large, but during most of the life time d(f) is of the order of 1. When non-uniformity on the grain size scale is investigated, l should be taken as RI. Substituting this, one obtains:

[G.24]

The term Lb/b in the right-hand side is the most important parameter since it relates the creep deformability of the grains to the diffusion parameter of the grain boundary cavitation. When Lb/b is very large, grain deformation is not expected to become significant so that cavitation is expected to be uniform. When Lb/b is very large, cavitation will be nonuniform. A grain size effect enters through the ratio b/RI.

2.5 Conclusion Real materials always contain different types of carbides, which leads to different equilibrium methane pressures. In general, the time to first cavity coalescence by HA is affected by the volume fractions of different carbides and their distributions inside the material. The main conclusion from the present study is that growth of all cavities is coupled in general by creep by virtue of similar mechanisms as first proposed by Dyson [9]. This leads to internal stresses that tend to constrain cavity growth in areas with above average methane pressures resulting from relatively unstable carbides present there. Evidently, the HA resistance of a material is enhanced by reducing the amount of unstable carbides; but, this study shows that the lifetime does not simple scale with the amount of unstable carbides. Unfortunately, a simple accurate estimate of the time to failure does not seem to be feasible in general. However, there are two limiting cases: (i) creep is so fast that creep constraints disappear; (ii) diffusion is so fast that constraint enforces uniform growth of all cavities irrespective of their internal pressure. In the first case, cavity evolution can be determined directly from cavity growth relations (5)–(10) by substituting the applied stresses. The second case allows for an estimate assuming rigid grains. A rough estimate whether or not the current HA conditions are a limiting case can be obtained through the value of the parameter defined in (16).


251

2.5.1 Acknowledgment The research of Marc van der Burg is sponsored by the Shell Research and Technology Center, Amsterdam, The Netherlands.


252

3.0 COMPARISON OF DELFT AND APPROACH REPORTED FROM JAPAN

3.1 Introduction to the Hydrogen Attack Models Hydrogen attack is a material degradation process occurring in steels exposed to high pressures of hydrogen at elevated temperatures. During the last twenty years, several models have been proposed to describe hydrogen attack [22,23,24]. Recently, Nomura and Sakai [25] presented a theoretical model for hydrogen attack (HA) at the ASME Pressure Vessel and Piping Conference 97, which has been picked up by API RP941. It is based on bubble growth due to diffusion. At the Delft University of Technology, a competing concept has been developed. There, the modeling is performed on three different length scales (cavity size [26], grain size [27] and specimen size scale [28]). The model on the smallest length scale, the so-called single cavity model, can be compared with the Nomura model. It is the aim of this report to point out the differences and the similarities of the Delft and the Nomura model and to compare some results. As already mentioned, the Delft model describes hydrogen attack on three levels. On the smallest length scale (microscopic level) the so-called single cavity model considers the growth of a single cavity due to grain boundary diffusion and creep driven by the methane pressure and the local stress state around the cavity. Analytical relations are found which can describe the void growth as a function of methane pressure, remote stress state, void radius, cavitation spacing, in terms of creep and diffusion parameters. These relations are then used in a model of an aggregate of grains (mesoscopic level). There, the effect of different methane pressure distributions on the damage evolution, on stress distributions etc. are studied with a finite element model. It turns out that two extreme cavitation modes occur. These two cavitation modes are covered by a continuum damage relation which can be used for the macroscopic modeling. In one extreme mode when the grains can deform significantly, the cavitation concentrates primarily on facets perpendicular to the macroscopic maximum principal stress. In the other mode, creep deformations of the grain are negligible and cavitation develops uniformly along all facets. The final continuum damage relation can be incorporated into a finite element model of, for example, a vessel to calculate the damage evolution, stress distribution and failure time of a vessel under internal hydrogen pressure. Nomura and Sakai‟s model [25] only addresses the microscopic level. They describe the growth of a single void with theoretical equations developed by Chuang et al. [22,29] which incorporate void growth by the diffusion of atoms from the surface to and along the grain boundaries. The comparison is structured as follows. We first compare those parts of the two models that address the calculation of the methane pressure from the hydrogen pressure. Then, by fixing the methane pressure, we proceed by comparing the void growth till coalescence as predicted by the two models when only diffusive processes are assumed. Finally we separately consider the contribution of creep deformation to void growth as is included in the Delft model but not in the Nomura model.


253

3.2 Calculation of the Methane Pressure Another difference between the two models originates from the calculation of the methane pressure. In the Nomura model the carbon activity is assumed for calculating the fugacity of methane according to . To evaluate and fit their parameters Nomura and Sakai [25] also conducted direct observation of void growth in two 2.25Cr-1Mo steels with a slightly different composition exposed to high pressure hydrogen. One tested steel contained a large amount of phosphor (steel P) while the other one contained a large amount of tin (steel Sn). In the modeling steel P got a carbon activity of 0.1 adapted from the data by Parthasarathy and Shewmon [24]. To obtain a good fit to the observed void growth the 2.25Cr-1Mo steel, with a large amount of tin (steel Sn) was taken to have a carbon activity of 0.04.

In the Delft model, the carbon activity is related to the contained carbides. The carbon activity is calculated with thermodynamic relations which depend on the type of carbides and the composition of carbides and matrix (26). Carbides found in 2.25Cr-1Mo steels are M3C, M2C, M7C3, M23C6, and M6C. In the Delft study, calculations have been carried out for the following carbides reacting with hydrogen to methane: (Cr0.3,Fe0.7)3C, (Cr0.6,Fe0.4)7C3, and (Cr0.5,Fe0.5)23C6, in a Fe-matrix with 2.3 at % Cr.

Knowing the fugacity of methane is not sufficient; the pressure of methane, , is required in the void growth relations (24) and (25). The Nomura and the Delft models use different relations to convert methane fugacity into methane pressure. Odette and Vagrali [31] developed a statistical mechanical-based high temperature and high pressure equation-of-state for methane. Based on this equation-of-state they proposed approximate solutions for three different fugacity ranges which are taken in the Delft model:

[G.25]

where:

[G.26]

The Nomura model relies on the following semi-empirical equation-of-state:

[G.27]


254

The following values of the constants provided the best fit, when is expressed in atm, R in atm 1/mol K and T in K (22):

[G.28]

3.3 Comparison Between Methane Pressures

It is interesting to study the influence of different carbon activities and different fugacity pressure relations on the resulting methane pressure for hydrogen pressures between 1 MPa and 100 MPa. 326HFigure G-20 shows the correlation between and one obtains at 400ºC and 600ºC by assuming the same carbon activity ( aC = 0.04, aC = 0.1) and fugacity-pressure relation as in the Nomura model (see 327HFigure G-21 of (25). Substituting the Nomura fugacity-pressure relation [G.27] by the Delft relation [G.26] leads to the -curves plotted in 328HFigure G-21. To get a clear picture of the influence of the fugacity-pressure relations the curves of 329HFigure G-20 and 330HFigure G-21 are put together in 331HFigure G-22 and restricted to realistic hydrogen pressures. It turns out that the fugacity-pressure relations used in the two models are responsible for different methane pressures and that they show a different temperature dependence. Having the same methane fugacity at 400ºC and applying the Delft fugacity-pressure relation leads to a higher methane pressure while at higher temperatures (600ºC) the Nomura relation predicts a higher methane pressure. 332HFigure G-23 depicts the correlation between methane pressure and hydrogen pressure given by the Delft model for the already described carbides M3C, M7C3 and M23C6. Once again the calculations of the methane pressure were repeated with the Nomura fugacity-pressure relation to see how the -curves related to the three carbide types change due to a different fugacity-pressure relation. The activities of the three carbides were the same as used in 333HFigure G-23. These resulting -curves are given in 334HFigure G-24


255

3.4 Description of Void Growth Due to Diffusion In the single cavity model of Delft the two deformation mechanisms are grain boundary diffusion and dislocation creep while the Nomura model only considers diffusion. Therefore, it only makes sense to compare the Nomura equation for void growth with that Delft equation which describes the void growth due to diffusion. The influence of creep on void growth will be studied separately in Section 6. In the Delft model following equation is used to compute the rate of the void radius only due to grain boundary diffusion

[G.29]

where the void radius, half void spacing and its ratio are represented by r,b and d(-r/b). Furthermore, Db and b are grain boundary diffusion coefficient and grain boundary thickness, respectively. S, , and k are surface energy, contact angle, atomic volume and Boltzmann constant. The total pressure pm is given by the sum of the partial pressures of methane and hydrogen (pm = + ). But for the purpose of a correct comparison we also neglect the hydrogen pressure just like Nomura and Sakai [25]. The derivation of formula (G.29) is outlined in [26]. It goes back to the Hull-Rimmer model [30] where two assumptions have been made. First, surface diffusion is presumed to be rapid enough so that the void retains a quasi-equilibrium spherical caps shape (grain boundary diffusion controlled). Secondly, the grains are assumed to be effectively rigid. The starting point in the Nomura model is an equation for void growth developed by Chuang et al. [8] for a thin and crack-like shaped void (see Eq. 1 of [25]). Such a shape occurs when the void growth is surface diffusion controlled. Nomura used a simplification of the void growth relation derived by Chuang et al. [29] (for more details see [22]):

[G.30]

DS and S are surface diffusion coefficient and surface thickness, respectively. Note the difference in the exponents of the pressure and the diffusion parameters in (G.29) and (G.30).


256

3.5 Comparison of Predicted Void Growth Due To Diffusion

In this section the Delft equation for void growth due to diffusion, (G.29), is applied to create the same diagrams as plotted in Nomura and Sakai [25]. Assuming the same methane pressures (Nomura carbon activities and Nomura fugacity-pressure relation (2)), the same surface energy (S = 1.95 J/m2), the same contact angle ( = 70º) and the same void half spacings b (b=5.6 m for steel P and b=2.2 m for steel Sn) as taken by Nomura and Sakai [25] allows for a direct comparison of the diffusion part of the Delft model and the Nomura model. 335HFigure G-25 shows the resulting void growth curves for the steels P and Sn exposed to a hydrogen pressure of 30 MPa and temperatures of 550ºC and 600ºC when we stick to the grain boundary diffusion parameters usually used in the Delft model [24,26]:

. [G.31]

336HFigure G-25 is an equivalent part to 337HFigure G-23 of Nomura and Sakai [25] which is reconstructed in 338HFigure G-26. It turns out that the Delft model predicts faster void growth for steel Sn than the Nomura model. The picture is not so clear for steel P, where we have higher internal pressures and a larger spacing between the voids. In the early stages, the Delft model also predicts a faster void growth than the Nomura model which changes at about 650h exposure times for 550ºC and 4500h for 600ºC. These differences arise due to the equations (3) and (4) which are the result of two different modes of void growth. Equation (3) describes one extreme mode when the voids maintain their rounded shape of uniform curvature while (4) is for the growth of a thin and crack-like void. One has to be aware of the fact that only the grain boundary diffusion parameters serve as input parameters in (G.29) while in (G.30) grain boundary and surface diffusion parameters are combined to To assure for a „fair‟ comparison between the two models, the same value should be taken for grain boundary diffusion in the Delft and the Nomura model. Nomura and Sakai [4] fitted the combined diffusion parameter D(T) to their experimental observation of bubble growth. Unfortunately, their given relation for

does not provide the value for Dbb which is important for a correct comparison and interpretation of 339HFigure G-25 and 340HFigure G-26. Therefore, we can only compare these two figures directly when we assume that the value taken for Dbb in the Delft model coincides with the value of Dbb in the Nomura model. This assumption leads to a

. So, the interpretation given in this paragraph for 341HFigure G-25 and 342HFigure G-26 is suitable for a material with = DSS/Db/b = 30 at 550ºC and = 16.5 at 600ºC. Which mode is active, is determined by the value of , by the void radius r and by the stress state.


257

Chuang et al. [29] estimated the ranges of validity for these two modes in case of uniaxial loading. For a given stress level, void growth in the quasi-equilibrium mode is favored, when surface diffusion is much more rapid than grain boundary diffusion (i.e. when is large) and also in the early stages of growth, when the void radius r and its ratio to the half spacing b are small. Conversely, the crack-like mode is favored, when is small and in later stages of growth. The predicted behavior of steel P corresponds with this. In early stages the quasi-equilibrium mode leads to a higher void growth while later the void will get a more crack-like shape (343HFigure G-25 and 344HFigure G-26). Further, the stress state also influences the actual mode. Lower stresses (in case of hydrogen attack: lower internal pressures) extend the validity of the quasi-equilibrium mode. This is the reason why steel Sn with its lower methane pressure remains in the quasi-equilibrium mode when we assume the aforementioned diffusion parameters. In the same way, we can compare the void growth of steel Sn at 525ºC and 550ºC without and with an applied stress of 100 MPa, 150 MPa and 200 MPa at a hydrogen pressure of 20 MPa predicted by the Delft model (quasi-equilibrium mode) with that predicted by the Nomura model (crack-like mode). Like Nomura and Sakai [25] we assume that the stress state remote from the void is equal to the applied stress state, although it has been shown with the Delft modeling on the second length scale that this is usually not the case [27]. 345HFigure G-27 and 346HFigure G-29 display the results of the Delft model ((G.29) and Delft diffusion parameters) for steel Sn under the described conditions. The applied stress is incorporated into (G.29) and (G.30) by replacing with + . 347HFigure G-28 and 348HFigure G-30 are reconstructions of 349HFigure G-29 (a) and (b) of Nomura and Sakai [25]. Despite of the higher stresses the quasi-equilibrium mode is still favored which can be seen in a faster void growth in 350HFigure G-27 and 351HFigure G-29 than in 352HFigure G-28 and 353HFigure G-30. Some further calculations were conducted with the Delft model and the Delft diffusion parameters. 354HFigure G-31 plots the exposure times at which the damage S (defined as the total cross section of voids per unit area of grain boundary) reaches 5% (solid line) or 10% (dashed line) at a hydrogen pressure of 30 MPa. This figure is a counterpart to 355HFigure G-26 of Nomura and Sakai [25].


258

3.6 Creep As An Additional Deformation Mechanism

As already pointed out, the Delft model cannot only describe the void growth due to diffusion but also due to dislocation creep. In contrast to the calculations before, we now allow for creep in the Delft model. Since the Nomura model does not cover creep, no comparison is possible. The objective of this section is to assess the importance of incorporating the creep distribution in the void growth model. The grains are assumed to deform by secondary creep with the effective creep strain rate

given by the Norton power-law

[G.32]

in terms of the creep exponent n, the reference stress parameter 0, the effective Mises stress o, and the temperature dependent reference strain rate parameter . The expression for the growth rate of the cavity radius r is now extended to have two contributions, i.e.

[G.33]

where the subscript df signifies the contribution by diffusion and cr is the contribution due to void growth in a creeping material. The creep contribution takes a rather wieldy form, which is not repeated here; full details may be found in [26, 28]. The diffusive contribution is essentially similar to (G.29) but with d2 being replaced by f . The latter is defined by

[G.34]

where:

[G.35]

This is an approximate, yet accurate way [32, 33] to incorporate the synergistic effect of creep on diffusive growth.


259

In reference to the creep parameters, it can be summarized that the following creep parameters have to be determined from creep tests: creep exponent n, activation energy Qv, reference creep strain rate , reference stress 0, and reference temperature 0. To study the influence of creep on the void growth, real creep parameters were taken from different sources. Obviously, creep depends on the steel composition and its microstructure, so that heat treatment has a strong influence. 356HTable G.1 lists the three different sets of creep parameters that were taken in the study. Data 1 are the data given by Frost and Ashby [34] for a 1%Cr-Mo-V steel. Data 2 and Data 3 stem from the Brite Euram Project PREDICH BRPR-CT96-0179 determined from creep tests of the 2.25Cr-1Mo standard steel and the modified 2.25Cr-1Mo steel. Since only the combination of is relevant and not and 0 separately, was determined by setting 0 equal to N/m2.

Table G-1 – Creep Parameters for Various Steels Used in Figure G-32 and Figure G-33

Set (s-1) 0 (N/m2) 0(ºC) Qv (kJ/mole) N Source

Data 1 5.7x10-62 1 455 251 6.0 Frost and Ashby Data 2 2.4x10-63 1 455 225 6.5 stand. 2.25Cr1Mo Data 3 2.5x10-53 1 455 380 5.0 mod. 2.25kCr-1Mo

The Delft model (including the creep part) is taken to calculate the void growth of a single void in steel P. The methane pressure is again computed according to the Nomura model (aC = 0.1, fugacity relation (G.27)), otherwise only Delft material parameters are used [5]. The creep parameters are taken according to Table G-1. 357HFigure G-32 shows the resulting void growth curves of steel P at 550ºC and 600ºC at 30 MPa hydrogen pressure in dependence of various creep data. Additionally, the void growth curves corresponding to grain boundary diffusion only are drawn in 358HFigure G-32 as solid lines. When the creep characteristics correspond to Data 1 [13], creep is too slow compared to diffusion to affect the void growth. With the steel creep properties similar to Data 2 or Data 3, the voids grow much faster, particularly during the later stages. At the beginning, void growth is dominated by diffusion which is confirmed by the effect that no matter which creep data are taken, the void growth curves coincide. But at later stages, the creep data can have a tremendous effect, which is obvious from comparing the curves related to Data 2 (dot-dashed lines) with the purely diffusive growth curves (solid lines). Having an additional uniaxial stress-state remote from the void should further increase the effect of creep due to its nonlinearity. This is indeed the case when one computes the void growth of steel P at 550ºC exposed to 20 MPa hydrogen pressure under a uniaxial stress of 150 MPa. 359HFigure G-33 shows the growth curves for various creep data (Data 1-3) together with the curves when only the methane pressure drives the growth ( = 0). One realizes that creep can accelerate void growth significantly. As mentioned before, this is an effect that is not incorporated in the Nomura model.


260

3.7 Conclusions The Delft model on the smallest length scale [26] describing void growth due to hydrogen attack was compared with the model developed by Nomura and Sakai [25]. The following differences were discovered:

1) To get the fugacity of methane, Nomura and Sakai [4] directly operate on an assumed carbon activity, while in the Delft model its activity is related to the microstructure of the steel through thermodynamic relations. Some uncertainties still remain in obtaining the actual carbon activity of a steel.

2) Different relations are used for converting the fugacity of methane into the partial pressure

of methane which lead to quite different results, particularly in the high range of fugacity. Further, the two relations describe a different temperature dependence, but 360HFigure G-22 shows that under typical hydrogen attack conditions this is of minor importance.

3) Both models consider void growth due to diffusion. But the used void growth relations differ

significantly because they describe different modes. In the Delft model, the voids maintain their quasi-equilibrium spherical caps shape (grain boundary diffusion controlled void growth) while in the Nomura model they are crack-like (surface diffusion controlled void growth). Diffusion parameters, stress state and void radius determine which mode will actually occur in reality. Unfortunately, it is difficult to determine the diffusion parameters experimentally. Therefore, it is desirable to have additional experimental evidence for the shape of the voids.

4) Only the Delft model can account for void growth due to creep. It could be shown that creep

can significantly accelerate void growth. Of course, this influence strongly depends on the actual creep and diffusion properties of the investigated steel.


261

4.0 FIGURES

FIGURE G-1 – GLOBAL COORDINATE SYSTEM OF THE UNIT CELL. DUE TO SYMMETRIES ONLY A QUARTER OF THE UNIT CELL WILL BE ANALYZED.

FIGURE G-2 – GRAIN BOUNDARY CAVITIES HAVING A DIAMETER OF 2A AND WHICH ARE SPACED WITH A DISTANCE OF 2B. THE GROWTH OF THE CAVITIES IS DRIVEN BY AN

INTERNAL PRESSURE OF PM , HOWEVER CAN BE ACCELERATED OR CONSTRAINED BY STRESSES REMOTE FROM THE CAVITY.


262

FIGURE G-3 – DUE TO CAVITY GROWTH THE ADJACENT GRAINS SEPARATE. THIS SEPARATION OF GRAINS CAN BE QUANTIFIED BY ‘SMEARING-OUT’ OF THE CAVITY VOLUME

OVER ITS FACET AREA.


263

FIGUREG-4 – FINITE ELEMENT MESH USED FOR THE NUMERICAL SIMULATIONS OF THE QUARTER UNIT CELL. EACH QUADRILATERAL CONSISTS OF FOUR TRIANGULAR

SUBELEMENTS. GRAIN BOUNDARY ELEMENTS ARE NOT SHOWN.


264

FIGURE G-5 – THE ASSUMED INTERNAL CAVITY PRESSURE Pm DISTRIBUTION ALONG THE GRAIN BOUNDARY FACETS FOR THE RESULTS SHOWN IN FIGS. 6–11 ( Pm IS PLOTTED

PERPENDICULAR TO THE FACETS), WITH A MAXIMUM VALUE P max m = 2.2Pm

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

265

FIG

UR

E G

-6 –

DA

MA

GE

EVO

LUTI

ON

FO

R T

HE

CA

VITY

PR

ESSU

RE

DIS

TRIB

UTI

ON

AC

CO

RD

ING

TO

FIG

UR

E 5

FOR

TH

REE

VA

LUES

OF

b 1 L

m .

FIG

UR

ES (A

)–(C

) SH

OW

TH

E D

AM

AG

E ST

ATE

IN T

ERM

S O

F a/

b (P

LOTT

ED N

OR

MA

L TO

, AN

D O

N E

ITH

ER

SID

E O

F, T

HE

GR

AIN

BO

UN

DA

RIE

S) F

OR

TH

E TH

REE

VA

LUES

OF

b 1 L

m W

HEN

(a/b

) max

= O

.3.

SIM

ILA

RLY

, (D

)–(F

) SH

OW

TH

E D

AM

AG

E ST

ATE

JU

ST P

RIO

R T

O F

IRST

CA

VITY

CO

ALE

SCEN

CE.

FIG

UR

ES (G

)–(I)

SH

OW

TH

E D

AM

AG

E EV

OLU

TIO

N IN

TIM

E FO

R T

WO

PO

INTS

ALO

NG

TH

E G

RA

IN F

AC

ET (S

OLI

D L

INES

). TH

E D

ASH

ED L

INES

SH

OW

TH

E SI

NG

LE C

AVI

TY M

OD

EL

RES

ULT

S.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

266

FIG

UR

E G

-7 –

DIS

TRIB

UTI

ON

S O

F (A

) TH

E G

RA

IN S

EPA

RA

TIO

N

, NO

RM

ALI

ZED

BY

THE

CU

RR

ENT

MA

XIM

UM

VA

LUE

max

=

3.0X

10-2

R1 ,

AN

D (B

) TH

E N

OR

MA

L ST

RES

S

n / P

m J

UST

PR

IOR

TO

FIR

ST C

AVI

TY C

OA

LESC

ENC

E FO

R L

OG

(b1 L

m )

= -0

.5.

THE

DA

RK

ER A

REA

S IN

(B) I

ND

ICA

TE T

HA

T TH

E N

OR

MA

L ST

RES

S IS

CO

MPR

ESSI

VE.

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

267

FIG

UR

E G

-8 –

EVO

LUTI

ON

OF

(A) T

HE

SEPA

RA

TIO

N R

ATE

, NO

RM

ALI

ZED

WIT

H T

HE

INIT

IAL

FAC

ET H

ALF

-WID

TH R

1 AN

D T

HE

CR

EEP

RA

TE P

AR

AM

ETER

p (

CF.

(13)

), A

ND

(B) T

HE

NO

RM

AL

STR

ESS

n / P

m A

T TH

E TR

IPLE

PO

INT =

0 A

ND

TH

E C

ENTE

R O

F TH

E FA

CET

=

0.5

FO

R T

HE

CA

SE W

HER

E (b

1 Lm

) = -0

.5 (

CF.

FIG

UR

E 7)

.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

268

FIG

UR

E G

-9 –

DIS

TRIB

UTI

ON

OF

(A) T

HE

SEPA

RA

TIO

N

/m

ax ,

WH

ERE

max

= 2

.7 X

10-2

R1,

AN

D (B

) TH

E N

OR

MA

L ST

RES

S

nPm

JU

ST P

RIO

R T

O F

IRST

CA

VITY

CO

ALE

SCEN

CE

FOR

LO

G (b

1 Lm

) = 0

.5 T

HE

DA

RK

ER A

REA

SIN

(B) I

ND

ICA

TE T

HA

T TH

E N

OR

MA

L ST

RES

S IS

CO

MPR

EHEN

SIVE

.

FIG

UR

E G

-10

– EV

OLU

TIO

N O

F (A

) TH

E SE

PAR

ATI

ON

RA

TE A

ND

(B) T

HE

NO

RM

AL

STR

ESS

AT

THE

TRIP

LE P

OIN

T =

0 A

ND

TH

E C

ENTE

R O

F TH

E FA

CET

=

0.5

FO

R T

HE

CA

SE W

HER

E LO

G (b

1 Lm

) = 0

.5.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

269

FIG

UR

E G

-11

– D

ISTR

IBU

TIO

N O

F TH

E G

RA

IN S

EPA

RA

TIO

N

, N

OR

MA

LIZE

D B

Y TH

E C

UR

REN

T M

AXI

MU

M V

ALU

E

max

= 3

.1

X 10

-2 R

1, JU

ST P

RIO

R T

O F

IRST

CA

VITY

CO

ALE

SCEN

CE

FOR

(b1 L

m) =

1.5

.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

270

FIG

UR

E G

-12

– (A

) TH

E A

SSU

MED

SH

OR

T W

AVE

LEN

GTH

, IN

TER

NA

L C

AVI

TY P

RES

SUR

E D

ISTR

IBU

TIO

N O

VER

TH

E PO

LYC

RYS

TAL.

(B) T

HE

SEPA

RA

TIO

N D

ISTR

IBU

TIO

N

/m

ax, W

HER

E

max

/ R

1 =

3.1

X 10

-2, F

OR

LO

G (b

1 Lm

) = 0

.5 J

UST

PR

IOR

TO

FIR

ST C

AVI

TY C

OA

LESC

ENC

E.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

271

FIG

UR

E G

-13

– IN

TER

NA

L C

AVI

TY P

RES

SUR

E D

ISTR

IBU

TIO

N O

VER

TH

E Q

UA

RTE

R U

NIT

CEL

L. O

NLY

TH

E C

AVI

TIES

ALO

NG

TH

E C

ENTR

AL

FAC

ET A

RE

PRES

SUR

IZED

. ON

TH

E O

THER

FA

CET

S TH

E IN

TER

NA

L C

AVI

TY P

RES

SUR

ES A

RE

ASS

UM

ED T

O

BE

ZER

O.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

272

FIG

UR

E G

-14

– (A

)–(C

): TH

E D

AM

AG

E ST

ATE

a/b

IN T

ERM

S O

F FO

R T

HR

EE D

IFFE

REN

T VA

LUES

OF

b 1 /

L m J

UST

BEF

OR

E FI

RST

CA

VITY

CO

ALE

SCEN

CE.

TH

E EV

OLU

TIO

N O

F TH

E D

AM

AG

E A

T TH

E FA

CET

MID

POIN

TS A

, B A

ND

C (S

EE F

IGU

RE

13)

AR

E SH

OW

N IN

(D)–

(F).

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

273

FIG

UR

E G

-15

– D

ISTR

IBU

TIO

N O

F (A

)–(C

) TH

E SE

PAR

ATI

ON

/

max

AN

D (D

)–(F

) TH

E N

OR

MA

L ST

RES

S

n / P

m O

VER

TH

E PO

LYC

RYS

TAL

FO

R D

IFFE

REN

T b 1

/ L m

(C

F. F

IGU

RE

14).

THE

MA

XIM

UM

SEP

AR

ATI

ON

S

max

AR

E (A

)–(B

) 3.1

X 1

0-2 R

1 , (C

) 2.

9 X

10-2

R1 .

AT

THE

DA

RK

ER A

REA

S IN

(D)–

(F),

THE

NO

RM

AL

STR

ESS

IS C

OM

PRES

SIVE

.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

274

FIG

UR

E G

-16

– EV

OLU

TIO

N O

F TH

E SE

PAR

ATI

ON

RA

TES

AN

D T

HE

NO

RM

AL

STR

ESS

AT

A, B

AN

D C

(SEE

FIG

UR

E 13

) FO

R

THE

DIF

FER

ENT

CA

SES

SHO

WN

IN F

IGU

RES

14–

15.


275

FIGURE G-17 – CLOSE UP OF A TRIPLEPOINT WHEN THE CREEP DEFORMABILITY OF THE GRAINS IS NEGLIGIBLE: THE SEPARATION RATES i AND j ON ADJACENT FACETS ARE

KINEMATICALLY COUPLED.

Rep

ort t

o A

PI -

The

Tech

nica

l Bas

is F

or A

PI R

P 94

1

276

FIG

UR

E G

-18

– (A

) SC

HEM

ATI

C M

OTI

VATI

ON

FO

R T

HE

TWO

–BA

R M

OD

EL A

LON

G A

GR

AIN

BO

UN

DA

RY

FAC

ET. (

B) T

HE

TWO

–BA

R M

OD

EL: A

PA

RA

LLEL

AR

RA

NG

EMEN

T O

F TW

O C

OU

PLED

BA

RS

WIT

H D

IFFE

REN

T C

AVI

TATI

ON

PR

OPE

RTI

ES.

REP

OR

T TO

API

- TH

E TE

CH

NIC

AL

BA

SIS

FOR

API

RP

941

277

FIG

UR

E G

-19

– (A

) VA

RIA

TIO

NS

IN T

HE

INTE

RN

AL

CA

VITY

PR

ESSU

RE

ALO

NG

A G

RA

IN B

OU

ND

AR

Y FA

CET

DU

E TO

D

IFFE

REN

CES

IN C

AR

BID

E R

EAC

TIVI

TY, (

B) L

EAD

ING

TO

A N

ON

-UN

IFO

RM

SEP

AR

ATI

ON

RA

TE D

ISTR

IBU

TIO

N W

ITH

G

RA

DIE

NT T

O B

E A

CC

OM

MO

DA

TED

BY

THE

AD

JAC

ENT

GR

AIN

MA

TER

IAL,

AN

D (C

) TH

E R

ESU

LTIN

G R

EAC

TIVE

NO

RM

AL

STR

ESSE

S O

N T

HE

GR

AIN

BO

UN

DA

RY.


278

FIGURE G-20 – CORRELATION BETWEEN METHANE PRESSURE AND HYDROGEN PRESSURE OBTAINED WITH NOMURA CARBON CAVITY AND NOMURA FUGACITY-PRESSURE RELATION

(USED IN NOMURA MODEL).


279

FIGURE G-21 – CORRELATION BETWEEN METHANE PRESSURE AND HYDROGEN PRESSURE OBTAINED WITH NOMURA CARBON ACTIVITY AN DELFT FUGACITY–PRESSURE RELATION.


280

FIGURE G-22 – INFLUENCE OF THE FUGACITY-PRESSURE RELATIONS ON THE METHANE PRESSURE BY ASSUMING NOMURA CARBON ACTIVITIES.


281

FIGURE G-23 – CORRELATION BETWEEN METHANE PRESSURE AND HYDROGEN PRESSURE OBTAINED FOR THREE DIFFERENT TYPES OF CARBIDES WITH DELFT FUGACITY–PRESSURE

RELATION.


282

FIGURE G-24 – CORRELATION BETWEEN METHANE PRESSURE AND HYDROGEN PRESSURE OBTAINED FOR THREE DIFFERENT TYPES OF CARBIDES WITH NOMURA FUGACITY–

PRESSURE RELATION.


283

FIGURE G-25 – VOID GROWTH OF THE STEELS P AND SN CALCULATED BY THE DIFFUSIVE PART OF DELFT MODEL AND THE DELFT DIFFUSION PARAMETERS AT A HYDROGEN

PRESSURE OF 30 MPA.


284

FIGURE G-26 – VOID GROWTH OF THE STEELS P AND SN CALCULATED BY THE NOMURA MODEL AT A HYDROGEN PRESSURE OF 30 MPA


285

FIGURE G-27 – VOID GROWTH OF STEEL SN AT 525o" C WITH AND WITHOUT STRESS

CALCULATED BY THE DIFFUSIVE PART OF THE DELFT MODEL AND THE DELFT DIFFUSION PARAMETERS AT A HYDROGEN PRESSURE OF 20 MPA.


286

FIGURE G-28 – VOID GROWTH OF STEEL SN AT 525oC WITH AND WITHOUT STRESS CALCULATED BY THE NOMURA MODEL AT A HYDROGEN PRESSURE OF 20 MPA


287

FIGURE G-29 – VOID GROWTH OF STEEL SN AT 550oC WITH AND WITHOUT APPLIED STRESS CALCULATED BY THE DIFFUSIVE PART OF THE DELFT MODEL AND THE DELFT DIFFUSION

PARAMETERS AT A HYDROGEN PRESSURE OF 20 MPA.


288

FIGURE G-30 – VOID GROWTH OF STEEL SN AT 550oC WITH AND WITHOUT STRESS CALCULATED BY THE NOMURA MODEL AT A HYDROGEN PRESSURE OF 20 MPA


289

FIGURE G-31 – EXPOSURE TIMES CAUSING THE SAME DAMAGE S (= VOID AREA PER GRAIN BOUNDARY) AT A HYDROGEN PRESSURE OF 30 MPA IN THE TEMPERATURE RANGE 400–600oC

CALCULATED BY THE DIFFUSIVE PART OF THE DELFT MODEL AND THE DELFT DIFFUSION PARAMETERS.


290

FIGURE G-32 – VOID GROWTH OF STEEL P CALCULATED BY THE DELFT MODEL (DIFFUSION + CREEP) AT A HYDROGEN PRESSURE OF 30 MPA IN DEPENDENCE OF THE CREEP

PARAMETERS LISTED IN TABLE G-1. THE SOLID LINES REFER TO VOID GROWTH ONLY DUE TO GRAIN BOUNDARY DIFFUSION.


291

FIGURE G-33 – VOID GROWTH OF STEEL P AT 550oC UNDER AN APPLIED STRESS OF 150 MPA AND WITHOUT STRESS, CALCULATED BY THE DELFT MODEL (DIFFUSION + CREEP) AT A

HYDROGEN PRESSURE OF 20 MPA IN DEPENDENCE OF THE CREEP PARAMETERS LISTED IN TABLE G-1. THE SOLID LINES REFER TO VOID GROWTH ONLY DUE TO GRAIN BOUNDARY

DIFFUSION.


292

5.0 REFERENCES

G.1 American Petroleum Institute Publication 941, (1990) 4th edition. G.2 M.W.D. van der Burg, E. van der Giessen and R.C. Brouwer, Acta mater., 44 (1996)505. G.3 H.-M. Shih and H.H. Johnson, Acta metall., 30 (1982) 537. G.4 T.A. Parthasarathy, Acta metall., 33 (1985) 1673. G.5 P.G. Shewmon, Acta metall., 35 (1987) 1317. G.6 A. Needleman and J.R. Rice, Acta metall., 28 (1980) 1315.M.W.D. van der Burg, E. van der Giessen / Cavitation-grain interactions during HA 19 G.7 T.-L. Sham and A. Needleman, Acta metall., 31 (1983) 919. G.8 E. van der Giessen, M.W.D. van der Burg, A. Needleman and V. Tvergaard, J. Mech. Phys. Solids, 43 (1995) 123. G.9 B.F. Dyson, Metal Science, 10 (1976) 349. G.10 V. Tvergaard, J. Mech. Phys. Solids, 32 (1984) 373. G.11 E. van der Giessen and V. Tvergaard, Acta metall. mater., 42 (1994) 959. G.12 M.W.D. van der Burg and E. van der Giessen, in S.I. Andersen et al. (eds.), 15th Risø International Symposium on Materials Science 1994, Risø National Laboratory, Roskilde, Denmark, 1994, p. 263. G.13 M.W.D. van der Burg and E. van der Giessen, in A.W. Thompson and N.R. Moody (eds.), 5th Int. Conf. on The Effects of Hydrogen on Material Behavior, Moran, Wyoming, 1994, TMS, Warrendale, PA, 1994, p. 313. G.14 G.R. Odette and S.S. Vagarali, Metall. Trans. A, 13A (1982) 299. G.15 T.A. Parthasarathy and P.G. Shewmon, Metall. Trans. A, 15A (1984) 2021. G.16 P.G. Shewmon, Mat. Sci. Tech., 1 (1985) 1. G.17 D. Hull and D.E. Rimmer, Phil. Mag., 4 (1959) 673. G.18 A.C.F. Cocks and M.F. Ashby, Progress in Materials Science, 27 (1982) 199. G.19 E. van der Giessen and V. Tvergaard, Mech. Mate., 17 (1994) 47. G.20 B. Budiansky, J.W. Hutchinson and S. Slutsky, in H.G. Hopkins and M.J. Sewell (eds.),Mechanics of Solids: The Rodney Hill 60th Anniversary Volume, Pergamon Press, Oxford, 1982, p. 13. G.21 B.F. Dyson, Revue Phys. Appl., 23 (1988) 605. G.22 G. Sundarajan and P.G. Shewmon, “The Kinetics of Hydrogen Attack of Steels,” Metall. Trans. 12A, p. 1761, 1981. G.23 H.-M. Shih and H.H. Johnson, “A Model Calculation of the Nelson Curves for Hydrogen Attack,” Acta Metall. 30, p. 537, 1982. G.24 T.A. Parthasarathy, “Mechanisms of Hydrogen Attack of Carbon and 2.25Cr-1Mo Steels,” Acta Metall. 9, p. 1673, 1985. G.25 T. Nomura and T. Sakai, “Theoretical and Parametric Methods for Evaluating Hydrogen Attack Limits,“ ASME Pressure Vessel & Piping Conference 97, July 27-31, 1997, Orlando, Florida. G.26 M.W.D. van der Burg, E. van der Giessen and R.C. Brouwer, “Investigation of Hydrogen Attack in 2.25Cr-1Mo with a High Triaxiality Void Growth Model,” Acta Mat. 44, p. 505,1996.


293

G.27 M.W.D. van der Burg and E. van der Giessen, “Non-Uniform Hydrogen Attack and the Role of Interaction with Creep,” Mat. Sci. and Eng. A220, p. 200, 1996. G.28 M.W.D. van der Burg, E. van der Giessen and V. Tvergaard, “A Continuum Damage Analysis of Hydrogen Attack in A 2.25Cr-1Mo Pressure Vessel,” Mat. Sci. and Eng. A, in print. G.29 T.-J. Chuang, K.I. Kagawa, J.R. Rice and L.B. Sills, “Non-Equilibrium Models for Diffusive Cavitation of Grain Interfaces,” Acta Metall. 30, p. 265, 1979. G.30 D. Hull and D.E. Rimmer, “The Growth of Grain-Boundary Voids Under Stress,” Phil. Mag. 4, p. 673, 1959. G.31 G.R. Odette and S.S. Vagrali, “An Equation-of-State for Methane for Modeling Hydrogen Attack in Ferritic Steels,” Metall. Trans. 13A, p. 299, 1982. G.32 A. Needleman and J.R. Rice, “Plastic Creep Flow Effects in the Diffusive Cavitation of Grain Boundaries,” Acta Metall. 28, p. 1315, 1980. G.33 T. L. Sham and A. Needleman, “Effects of Triaxial Stressing on Creep Cavitation of Grain Boundaries,” Acta Metall. 31, p. 919, 1983. G.34 H.J. Frost and M.F. Ashby, “Deformation-Mechanism Maps,” Pergamon Press, Oxford, p.62, 1982.17.


294

APPENDIX H – SAMPLE PROBLEMS UTILIZING APPENDIX C AND DISCUSSION DEVELOPED BY SOME

COMMITTEE MEMBERS DURING REVIEW OF THE REPORT


295

1.0 ESTIMATING DAMAGE RATES FOR LIFE ASSESSMENT: BASIC METHODOLOGY

Cautionary Note from API: The following methodology is the product of a 941 Task Group subteam that reviewed the technical basis for the 941 Recommended Practice. The methodology presented has NOT been fully peer-reviewed, and is NOT to be considered as part of the 941 Recommended Practice. The reader is especially cautioned to consider the warnings regarding operating conditions and metallurgical factors noted below. It is included here to elaborate on the implementation of the approach described in Appendix C.

1.1 The basic approach to determine the likelihood of damage in HTHA service is: 5) Determine whether the methane formation rate is high enough that HTHA is a potential

concern. 6) If the methane formation rate is high enough to be a potential concern, then use methane

pressure vs. creep strength relations to determine if the damage is likely to be significant within the time of interest.

1.2 The specific steps are:

Step 1: Find your hydrogen pressure and temperature conditions on Figure C-1, “Reaction Rate Limits”. Note this figure gives a series of curves below which reaction rates are estimated to be too slow for attack to proceed. If the process conditions are below the curve for a material, then no HTHA is predicted and the analysis is done. If it is above the curve, go to Step 2.

7) The curves correspond to (in order):

100% Ordinary carbon steel 50% C-1/Mo Susceptible 25% C-1/Mo Less Susceptible 0.13% More susceptible 1 Cr- ½ Mo or 11/4 Cr- - ½ Mo -Si 0.10% Less Susceptible 1 Cr- ½ Mo or 1/1/4 Cr - ½ Mo- Si 0.08% High Strength 2 1/4 Cr 1 Mo 0.05% Conventional 2 1/4 Cr 1 Mo 0.03% 3Cr/2 1/4CrV modified

8) For carbon steel, one can get an ever better feel for the rate of methane formation by

examining Figure C-2. This figure tells you roughly the rate of methane formation relative to the “base” of carbon steel at the Nelson Curve limits (in which methane formation is considered to be so slow as to be no practical concern for any vessel life). Carbon steel is the only material for which we have such a curve, but it does not stop us from making estimates for other materials discussed in the API RP 941.

Step 2: If the case “fails” Step 1, then find the appropriate methane pressure vs creep strength curve (see below). Locate your temperature and hydrogen partial pressure to find the methane pressure and the estimated rate of hydrogen-attack damage.


296

9) The damage rates are given in the form of (for example) 10-6/hr. The rates are indexed to give you a million hour “life” (for the 10-6/hr rate), 100,000 hour life, and 10,000 hour life. Most equipment that operates indefinitely has creep rates at about 1-3 X 10-8/hr or lower. These rates assume that the methane generation rate is high enough to fill the voids faster than the voids will expand due to creep.

10) In some cases, if the methane formation rate is low, damage rate estimated from the methane pressure vs. creep strength curves used in Step 2 are probably conservative. One of the example cases shown below has this situation. Another element of conservatism has to do with the fact that damage calculated is local to the pressure used in the calculation and through wall propagation must be considered.

Step 3: The final step is to compare the expected “life” rate with the time of interest. If the expected “life” is not considerably more than the time interval of interest (for example, until the next opportunity for equipment replacement or detailed inspection), then the equipment would be considered at risk. Caution: The curves are necessarily generalized and have inherent inaccuracies. In addition errors introduced by the inaccuracies in knowing the temperature and hydrogen partial pressure, the actual stress, and the precise metallurgical condition can have a major effect upon the assessment. The methodology and the curves must be used with great caution.

2.0 EXAMPLES: HYPOTHETICAL CASES USING APPENDIX C

2.1 Case 1: Carbon Steel channel flange and shell welds discovered in 1.25Cr exchanger

During the RBI review of a hydroprocessing unit, it was discovered that the channel flange and long seam welds for the reactor feed/effluent exchanger were carbon steel.

Operating Conditions Life 34 years Hydrogen pressure 225 psig

Operating Temperature Case 1A 620 max Case 1B 750 max

A review of the past operating history developed the two sub-cases given below to bound the likely actual process conditions. No HTHA inspections have ever been performed on this piece of equipment because the belief was that the exchanger was 1 1/4 Cr. alloy and operating in a range where it would be susceptible to HTHA under operating conditions.

2.2 Case 1A: 225 psia at 620oF

Figure C-1 shows that you are above the 100% or carbon steel line. This means there could be a sufficient rate of methane formation for HTHA .


297

Step 2: Go to Figure C-2 to find that conditions are at about the 2X line. This means that the rate of methane formation is twice the “base” CS line (for which we have no concern with HTHA), but it is still a low rate of methane formation.

Step 3: Go to Figure C-3 (methane pressure vs. creep strength curves for CS). Conditions are not far above the 10-4 /hr curve. This is a fairly complex case because there is potentially methane pressure to increase the creep rate. But only a 2X methane formation rate may be marginal with regard to methane formation. The acceleration of the creep rate will not be as bad as we might think. Therefore it should be conservative to predict that we should not have severe creep damage before 10,000 hours, but there could be severe damage at 300,000 hours. Because of the desired 30 year service, it should be high priority to deal with this unit.

2.3 Case 1B: 225 psia at 750oF

Following the procedure above, we find the methane formation rate is 8X to 16X at the higher temperature. Therefore, in this case the methane formation rate may not be limiting and one could certainly expect damage within the time frame. Case 1A is of significant concern, Case 1B is viewed as substantially worse.

2.4 Case 2: Carbon steel pipe

Carbon steel piping operating for 20 years at 700F and 80 psia hydrogen partial pressure, not stress relieved, stresses within governing code allowable limits.

Step 1: As seen in Figure C-1, conditions are close to the 50% line. This says for carbon steel (100% line) there could be significant rate of methane formation. Step 2: Since this is carbon steel, we can go to Figure C-2 to see more closely that it is between 2X and 4X. This shows the possibility of attack, at a slow, but operationally significant rate. Go to Step 3. Step 3: For 700oF and 80 psia, per Figure C-3 the damage rate is close to the 10-6 /hr line, so this should not be a problem, even for 20 year old piping. However, this analysis must be done very cautiously because the methane reaction rate is significant and piping stresses can greatly increase the potential for local hydrogen damage. At a minimum, some inspection and metallurgical analysis would be warranted, which due to the localized nature could be difficult to find. Many operators choose to replace such piping rather than to try to keep ahead of damage through inspection.

2.5 Case 3: C-0.5Mo exchanger channel cover


298

An existing exchanger channel cover (A 204 Grade B, bare) has been operating for 30+ years. The inlet process conditions are 640 psia hydrogen at 680oF. The outlet conditions are 640 psia hydrogen partial pressure at 540oF. The cover is operating within code allowable stresses and is 4.1875-inch thick. Step 1: Curve C-1 says that the 540oF condition should be marginally OK (just below the 50% (C-0.5Mo) curve which is intended to be appropriate for highly susceptible C-0.5Mo steel). This would be a low priority for inspection, no further analysis needed. This assessment is based upon the expectation that the steel has been normalized and otherwise displays an appropriate microstructure. In addition, there is confidence that operating conditions are well known. The 680oF condition is well above the 25% curve, suggesting that the rate of methane formation would be sufficient to attack optimal C-0.5Mo steel. Go to Step 2. Step 2: Go to Curve C-4 to determine the likely rate of damage for C-0.5Mo. Assuming that the methane pressure reaction rate is fast enough to fill the creep/hydrogen voids, we see that at 680oF and 640 psia hydrogen the attack damage rate is in the range of 10-6 to 10-5 /hr. Therefore there is a strong potential for damage eventually. It could take tens of thousands of hours. This would be a good candidate for further analysis to determine rate of progression of damage through the thickness of the material. Figure 47 and the text refers to this concept, the technology has been developed further within the C-1/2 Mo Joint Industry Project.

2.6 Case 4: C-0.5Mo Methanator Short-Time Excursion

A process runaway caused a short-term temperature excursion of the 8 foot diameter, 1.25-inch thick C-0.5Mo reactor vessel. The temperature reached a peak of 1100oF for two hours, with a time-average temperature of 950oF for 48 hours. It has now dropped to its normal operating temperature of 450oF. Hydrogen partial pressure throughout the event never changed from its normal operating conditions of 200 psia. The normal total operating pressure of 250 psia remained constant throughout the excursion. Step 1: Figure C-1 shows that at temperatures above about 650oF and the partial pressure noted, the rate of methane formation is rapid for this material. This case warrants further analysis. Step 2: Figure C-4 shows that at 950oF and 200 psia is close to the 10-6 /hr curve. A 48 hour event should not be a major concern, when followed , as in this case, by benign conditions. The 1100oF condition is beyond the curve, but conservative extrapolation suggests that it is unlikely that the10-4 /hr rate has been exceeded. So for the two hours the risk of measurable damage is very small. This event would raise more mechanical concerns than HTHA concerns.


299

2.7 Case 5: Mn-0.5Mo exchanger with austenitic weld overlay/clad

For the 23 years the following clad/overlay, Mn-0.5 Mo exchanger has operated with the process stream near the early design curves for 0.5Mo steel. Operating conditions are given below. In the future they will need to operate at 810oF. Data:

Component Material Base Plate Thickness

Clad Thickness

WOL Thickness

Operating Temp

H2

Pressure

Design Total

Pressure Design Temp

Channel SA 302 B Mn-0.5Mo 1.25 in. 0.1 in.

316L SS N/A 750oF now 810oF future

Case A

670 psia

Case B

810 psia

1100 psig 850oF

Channel Cover

assumed to be Mn-0.5Mo

6.4375 in. N/A 0.3 750oF now 810oF future

Case A

670 psia

Case B

810 psia

1100 psig 850oF

Shell SA 302 B Mn-0.5Mo 1.5 in. 0.1 in.

316L SS N/A 640oF 800 psia 1100 psig 750oF

2.8 Case 5A: Consider Continued Operation at Current Conditions Step 0: Check the benefit of the clad or weld overlay. For the case of the channel, the hydrogen partial pressure ratio per Figure 20 (Austenitic Stainless Steel Clad/Overlay) is about 1/3 giving an effective hydrogen pressure of only 225 psi. For the channel cover the benefit of the weld overlay is a factor of 0.5, so the effective hydrogen pressure is 335 psi. For the shell thickness of 1.5-inch with 0.1 in. clad the effective hydrogen pressure will be 40% of the full pressure, or about 325 psi. Step 1: Use Figure C-1 to check the rate of methane formation.

Channel 225 psi and 750oF Channel Cover 335 psi and 750oF Shell 325 psi and 640oF

For these purposes we assume the Mn-0.5Mo resistance is the same as C-0.5Mo. For all three components, Figure C-1 shows the rates are above the 25% curve (for good C-0.5 Mo). So we need to analyze further. Step 2: Use Figure C-4 to check the likely damage rate.


300

Using Figure C-4, we see that each case the projected damage rate is (just) below (slower) than the 10-6 /hr rate, so we should be OK. As a result, these components would be low inspection priorities.

2.9 Case 5B: Operations wants to increase the operating conditions

Using the same process it is seen that the unit would have an order of magnitude of increased exposure (risk). This is left for the reader to confirm.

2.10 Case 6: 1Cr catalytic reformer reactor

A 1Cr (Gr12, not 1.25Cr) catalytic reformer reactor leaked on-stream at a top inlet nozzle. It went into service in 1952. In addition to cracking, there were blisters in the damaged areas (some visible from the ID). Assume the process conditions throughout the reactor life were:

930oF - 935oF inlet (at the top nozzle) 822oF outlet 325 psia hydrogen partial pressure

Also, this plant cycled up and down approximately monthly up for at least the last few years. We will examine the worst set of conditions, which is the inlet nozzle. Step 1: Check methane formation rate, Go to C-1. At 935oF and 325 psia conditions are between the 0.13% and 0.10% curves (that is, it is within the “band” of 1Cr/1.25Cr good and bad material). This requires further analysis. Step 2: Without the damage rate curves for 1Cr or 1.25Cr one may go instead to Figure 60 which shows that at about 325 psia hydrogen pressure, the methane pressure will be only about 2 ksi, and is well below the long term strength curve shown there. Therefore, absent significant residual or thermal stresses, one would not predict HTHA for this material.

2.11 Case 7: Over temperature Conditions in a 2.25Cr 1 Mo Reactor

2.25Cr hydroprocessing reactor vessel, designed to ASME Div. 2, operated within allowable stresses, at 840oF with 2200 psia hydrogen partial pressure for 30 years. In addition, it accumulated 6 months at 900oF metal temperature at the same hydrogen partial pressure. Reactor was returned to returned to 840oF. Assume first base conditions, no excursion.


301

Step 1: Go to C-1, note that normal operating conditions are slightly above the .05% curve, which is for ordinary 2.25Cr 1 Mo steel. (Q&T, higher strength, low PWHT is the 0.08% curve, conventional material tempered at 1275oF or higher is appropriate to the 0.05% curve). Closer analysis is needed. Step 2: Go to C-5 for 2.25Cr material. At 2200 psia and 840oF, you see that you between 10-5 and 10-6 rate. If you were truly operating continuously at 840oF, you might have problems. Many such vessels have clad and overlay. While the benefit of austenitic clad or overlay diminishes with backing steel thickness, in this case even a 20% pressure reduction would move the operating conditions below the threshold of concern. This vessel would require further study and involvement by materials engineers given its borderline potential. Now consider the excursion. Step 2A: 900oF for 4,000-5000 hours. Looking at Figure C-6, the damage rate exceeds 10-5 /hr at 900oF and is probably 3-4x greater than at the design condition. So the excursion would have used perhaps 20,000 equivalent hours at the normal operating condition. However, an additional concern is that the applied stress due to the process pressure would have exceeded the allowable stress at the higher temperature by as much as 40%, This might contribute greatly to the cavity nucleation and growth rates. The pressure of methane in existing voids would be increased above equilibrium, while the creep resistance of surrounding material would be reduced. Any such excursion must be treated extremely conservatively. Note: this case is significantly more severe than how our industry operates hydroprocessing reactors. However, it shows that: Short term excursions do not necessarily cause HTHA of 2.25Cr reactors, even if not clad. On the other hand, for continuous operation at (say) 825oF, the curves would predict normal

design lives, but conditions are not be far from predicting a significant risk of damage for 2.25Cr. Certainly operating at 840oF would be cause for close study.

Invoice To (❏ Check here if same as “Ship To”)

Name:

Title:

Company:

Department:

Address:

City: State/Province:

Zip/Postal Code: Country:

Telephone:

Fax:

Email:

❏ Payment Enclosed ❏ P.O. No. (Enclose Copy)

❏ Charge My IHS Account No.

❏ VISA ❏ MasterCard ❏ American Express❏ Diners Club ❏ Discover

Credit Card No.:

Print Name (As It Appears on Card):

Expiration Date:

Signature:

Quantity Title Total

Subtotal

Applicable Sales Tax (see below)

Rush Shipping Fee (see below)

Shipping and Handling (see below)

Total (in U.S. Dollars)

★ To be placed on Standing Order for future editions of this publication, place a check mark in the SO column and sign here:

Pricing and availability subject to change without notice.

Date:

SO★ Unit Price

❏ API Member (Check if Yes)

Ship To (UPS will not deliver to a P.O. Box)

Name:

Title:

Company:

Department:

Address:

City: State/Province:

Zip/Postal Code: Country:

Telephone:

Fax:

Email:

Mail Orders – Payment by check or money order in U.S. dollars is required except for established accounts. State and local taxes, $10 processing fee, and 5% shipping must be added.Send mail orders to: API Publications, IHS, 15 Inverness Way East, c/o Retail Sales, Englewood, CO 80112-5776, USA.

Purchase Orders – Purchase orders are accepted from established accounts. Invoice will include actual freight cost, a $10 processing fee, plus state and local taxes.Telephone Orders – If ordering by telephone, a $10 processing fee and actual freight costs will be added to the order.Sales Tax – All U.S. purchases must include applicable state and local sales tax. Customers claiming tax-exempt status must provide IHS with a copy of their exemption certificate.Shipping (U.S. Orders) – Orders shipped within the U.S. are sent via traceable means. Most orders are shipped the same day. Subscription updates are sent by First-Class Mail. Other options, including next-day service, air service, and fax transmission are available at additional cost. Call 1-800-854-7179 for more information.Shipping (International Orders) – Standard international shipping is by air express courier service. Subscription updates are sent by World Mail. Normal delivery is 3-4 days fromshipping date.Rush Shipping Fee – Next Day Delivery orders charge is $20 in addition to the carrier charges. Next Day Delivery orders must be placed by 2:00 p.m. MST to ensure overnight delivery.Returns – All returns must be pre-approved by calling the IHS Customer Service Department at 1-800-624-3974 for information and assistance. There may be a 15% restocking fee.Special order items, electronic documents, and age-dated materials are non-returnable.

Effective January 1, 2008.API Members receive a 30% discount where applicable.The member discount does not apply to purchases made for the purpose of resale or for incorporation into commercial products, training courses, workshops, or othercommercial enterprises.

Available through IHS:Phone Orders: 1-800-854-7179 (Toll-free in the U.S. and Canada)

303-397-7956 (Local and International)Fax Orders: 303-397-2740Online Orders: global.ihs.com

2008 PublicationsOrder Form

API provides additional resources and programs to the oil and natural gas industry which arebased on API Standards. For more information, contact:

API MONOGRAM® LICENSINGPROGRAMPhone: 202-962-4791Fax: 202-682-8070Email: [email protected]

API QUALITY REGISTRAR(APIQR®)> ISO 9001 Registration> ISO/TS 29001 Registration> ISO 14001 Registration> API Spec Q1® RegistrationPhone: 202-962-4791Fax: 202-682-8070Email: [email protected]

API PERFORATOR DESIGNREGISTRATION PROGRAMPhone: 202-682-8490Fax: 202-682-8070Email: [email protected]

API TRAINING PROVIDERCERTIFICATION PROGRAM(API TPCPTM)Phone: 202-682-8490Fax: 202-682-8070Email: [email protected]

API INDIVIDUAL CERTIFICATIONPROGRAMS (ICP®)Phone: 202-682-8064Fax: 202-682-8348Email: [email protected]

API ENGINE OIL LICENSING ANDCERTIFICATION SYSTEM (EOLCS)Phone: 202-682-8516Fax: 202-962-4739Email: [email protected]

API PETROTEAM (TRAINING,EDUCATION AND MEETINGS)Phone: 202-682-8195Fax: 202-682-8222Email: [email protected]

API UNIVERSITYTM

Phone: 202-682-8195Fax: 202-682-8222Email: [email protected]

Check out the API Publications, Programs,and Services Catalog online at www.api.org.

Copyright 2008 – API, all rights reserved. API, API monogram, APIQR, API Spec Q1,API TPCP, ICP, API University and the API logo are either trademarks or registeredtrademarks of API in the United States and/or other countries.

THERE’S MOREWHERE THIS CAME FROM.

Additional copies are available through IHSPhone Orders: 1-800-854-7179 (Toll-free in the U.S. and Canada)

303-397-7956 (Local and International)Fax Orders: 303-397-2740Online Orders: global.ihs.com

Information about API Publications, Programs and Servicesis available on the web at www.api.org

1220 L Street, NWWashington, DC 20005-4070USA

202.682.8000

Product No. C09410

Date post:	27-Apr-2015
Category:	Documents
Upload:	verdiblue
View:	1,912 times
Download:	107 times

API - Technical Report 941 Sept. 2008 - The Technical Basis Document for API RP 941

Documents