Date post: | 28-Feb-2019 |
Category: |
Documents |
Upload: | hoangkhanh |
View: | 237 times |
Download: | 0 times |
1
A postA post--processor for fatigueprocessor for fatigue--crack growth analysis crack growth analysis based on a finitebased on a finite--element stress field and its element stress field and its
application to components with random defects application to components with random defects
ETH Zürich, 25 Oktober 2007Gunnar Härkegård, NTNU, Trondheim
und Zentrum für Mechanik, ETH
Vortrag im Rahmen desKOLLOQUIUMS FUER TECHNISCHE WISSENSHAFTEN
und desSEMINARS IN MECHANIK
2
AcknowledgementAcknowledgement
This presentation is largely based on work carried out at the Department of Engineering Design and Materials, NTNU, Trondheim, byArne Fjeldstad funded through the NorLight program and defending his PhD thesis 30/11, ‘Modelling of fatigue crack growth at notches and other stress raisers’Anders Wormsen funded by GE Energy and defending his PhD thesis 26/11, ‘A fatigue assessment methodology for notched components containing defects’
3
ContentsContents
BackgroundCracks growing from a single defectGrowth of short fatigue cracksRandom-defect analysisConclusions, outlook
4
BACKGROUNDBACKGROUND
5 Rolled material Cast material
Beachmarks of crack emanating from leading edge blade/ring transition (Huth)
6
Beachmarks of crack emanating from blade/ring transition (Huth)
7
Striations observed in an aluminium alloy Striations observed in an aluminium alloy (A. (A. JernbergJernberg, NTNU), NTNU)
8
In situ observation of crack growth by In situ observation of crack growth by means of SEM in an aluminium alloy means of SEM in an aluminium alloy (M. (M. AnderssonAndersson, LTH, , LTH, dissdiss. 2005). 2005)
Thickness = 0.8 mm, crack depth > 2 mm
9
‘‘A fatigue crack is growing from the very A fatigue crack is growing from the very first loading cycle.first loading cycle.’’Keith J. Miller, 1932Keith J. Miller, 1932--20062006
10
DCPD measurement of naturally initiated DCPD measurement of naturally initiated fatigue crack at notch root: Test specimen fatigue crack at notch root: Test specimen
11
DCPD measurement of naturally initiated DCPD measurement of naturally initiated fatigue crack at notch root: Test setupfatigue crack at notch root: Test setup
12
Crack growth at the notch root from the first Crack growth at the notch root from the first cycle! (K. cycle! (K. StStäärkrk))
D1392 Risswachstum an Kerben St572S, R0.5, RTEndrisslänge Kerbe A 0.84mm, Kerbe B 0.78mm
02468
10121416
0 500 1000 1500 2000
Zyklenzahl N
Pote
ntia
länd
erun
g (%
)
POT1(% ) POT2(% )
13
Growth of edge throughGrowth of edge through--cracks and cracks and semisemi--elliptical cracks in elliptical cracks in
inhomogeneous stress fieldsinhomogeneous stress fields
14
Growth of (large) fatigue cracks Growth of (large) fatigue cracks
15
Growth of a fatigue crack from Growth of a fatigue crack from aaii = 0.05 mm to= 0.05 mm toaaff = 5 mm in a = 5 mm in a homogeneoushomogeneous stress field stress field
0
0,5
1
1,5
2
2,5
3
3,5
4
4,5
5
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
N/N 0
a, m
m
Paris exponentm = 3
16
SemiSemi--elliptic surface crack (throughelliptic surface crack (through--crack crack has has a/ca/c = 0)= 0)
17
Surface crack in linearly decreasing stress Surface crack in linearly decreasing stress fieldfield
18
Normalised fatigue life as a function of the Normalised fatigue life as a function of the relative stress gradientrelative stress gradient
19
SemiSemi--elliptic crack at the root of a notchelliptic crack at the root of a notch
20
Surface crack at the root of a notchSurface crack at the root of a notch
21
Normalised fatigue life as a function of the Normalised fatigue life as a function of the curvature of the notch rootcurvature of the notch root
22
MODELLING THE GROWTH OFMODELLING THE GROWTH OFA FATIGUE CRACK FROMA FATIGUE CRACK FROM
A SINGLE DEFECTA SINGLE DEFECT
23
Different approaches to fatigue analysisDifferent approaches to fatigue analysis
Random DefectSingle DefectExplicit FCG analysisda/dn = f(Δσ, a; R),
Weakest LinkLocal StressImplicit FCG analysisS-N-curve (a > 1 mm),‘crack initiation’
ProbabilisticDeterministicMaterial properties
Crack Growth
24
Comprises all four approaches to fatigue designWritten in standard FORTRAN Can be operated under Windows and UNIX/LINUXCompatible with standard finite element
codes such as ABAQUS, ANSYS and I-DEAS
P•FAT – Probabilistic Fatigue Assessment Tool developed at
NTNU/IPM 2003-2007
25
P•FAT: Single Defect
Simulates growth of fatigue crack from crackSimulates growth of fatigue crack from crack--like defectlike defectArbitrary component geometrySingle crack-like defect of elliptical shape
• Embedded crack• Surface crack• Corner crack
Short-crack modelMean-stress correction (residual stress)Adaptive step-size controlUpdates location of crack front relative to free surface
26
Definition of crack plane and localDefinition of crack plane and localcoco--ordinate systemordinate system
27
Crack configurations implemented in Crack configurations implemented in PP••FATFAT
28
Computation of Computation of KKII(P) from stress field (FEA) (P) from stress field (FEA) and weight function (integration mesh)and weight function (integration mesh)
AyxgyxKA
z d)P,,(),()P(crack
I ′′⋅′′= ∫ ′σ
29
WeightWeight functionsfunctions for for embeddedembedded crackscracks
( ) 3 2 21 2 3 4
2', '; ' 18 8 8 8
s s s s sg x y Pπ ρ ρ ρ ρ ρ
= − − − −
Ref: Wang et al., Engineering Fracture Mechanics 59(3):381-392, 1998
( ) ( )els Gauss Gauss
i
N
i j j1 1 1
, , ; 'N N
k i j
K g Pσ ξ η ξ η= = =
⎛ ⎞≈ ⎜ ⎟
⎝ ⎠∑ ∑ ∑
30
KK estimationestimation accuracyaccuracy ((emdeddedemdedded crackscracks))
0
KFaσ π
=( ) 0''
iyya
σ σ ⎛ ⎞= ⎜ ⎟⎝ ⎠
31
WeightWeight functionsfunctions for for semisemi--ellipticelliptic crackscracks
( ) ( )( )( )
AA
2 1 '', / , /''; /
2 ''
f y a c a Lg y a c
a yπ
+=
−( ) ( )( )C
C
2 1 '', / , /''; /
''
f y a c a Lg y a c
yπ
+=
Ref: Shen and Glinka, Theoretical and Applied Fracture Mechanics 15:247-255, 1991
32
KK estimationestimation accuracyaccuracy ((semisemi--ellipticelliptic crackscracks))
0
KFaσ π
=( ) 0''
iyya
σ σ ⎛ ⎞= ⎜ ⎟⎝ ⎠
331
2
3
Smooth blockSmooth block--shaped specimen under axial shaped specimen under axial pushpush--pull loading, pull loading, RR = = --11
34
FCG through crossFCG through cross--section of smooth section of smooth blockblock--shaped specimenshaped specimen
35
Fatigue lives of cracked smooth specimens Fatigue lives of cracked smooth specimens based on Pbased on P••FAT and an eigenstrain methodFAT and an eigenstrain method
1027711419422
1588016164221
2018822901211
P•FATDai et al.Lciai
Cycles until failureInitial crack geometry (mm)
Dai et al. studied the growth of near surface cracks in a wide body by using a solution for the stress intensity factor based on eigen-strains.
Ref: Dai et al. Engineering Fracture Mechanics, 59(4):415-424, 1998
36
1
2
3
Notched blockNotched block--shaped specimen under shaped specimen under axial pushaxial push--pull loading, pull loading, RR = = --11
37
FCG through crossFCG through cross--section of notched section of notched blockblock--shaped specimenshaped specimen
38
Fatigue lives of cracked notched wide plate Fatigue lives of cracked notched wide plate based on Pbased on P••FAT and an approximate FAT and an approximate
analytical methodanalytical method
39
GROWTH OF SMALL FATIGUE CRACKSGROWTH OF SMALL FATIGUE CRACKS
40
KitagawaKitagawa--Takahashi diagram and Takahashi diagram and El Haddad intrinsic crack length modelEl Haddad intrinsic crack length model
for steel (for steel (RRp0.2p0.2 = 200 = 200 –– 800 800 MPaMPa), Cu and Al), Cu and Al
41
Generalisation of El HaddadGeneralisation of El Haddad’’s model s model for the fatigue limitfor the fatigue limit
‘‘IntrinsicIntrinsic’’ crack length:crack length:
42
ModifiedModified KK--T diagram T diagram
43
KK--T T ‘‘Fatigue Assessment DiagramFatigue Assessment Diagram’’
44
El HaddadEl Haddad’’s model in terms of s model in terms of ΔΔKKand and ΔΔσσ::
2
A
2th
eff 1 ⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
⎟⎠⎞
⎜⎝⎛
ΔΔ
+Δ=Δσσ
KKKK
Stress intensity range corrected Stress intensity range corrected with respect to (short) crack length:with respect to (short) crack length:
45
‘‘EquivalentEquivalent’’ stress of a crack in an stress of a crack in an inhomogeneous stress fieldinhomogeneous stress field
46
ddaa/d/dnn vs. vs. ΔΔKK for short cracks in AA6082for short cracks in AA6082--T6 T6 with the crack depth with the crack depth aa as a parameteras a parameter
1.E-12
1.E-11
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
1.E-05
1 10
ΔK , MPa m1/2
da/d
n, m
/cyc
le
a = 0.01 mma = 0.02 mma = 0.05 mma = 0.1 mma = 0.2 mma = 0.5 mma = 1 mma = 10 mm
47
ddaa/d/dnn vs. vs. ΔΔKK for short cracks in AA6082for short cracks in AA6082--T6 T6 with the stress range with the stress range ΔΔσσ as a parameteras a parameter
1.00E-11
1.00E-10
1.00E-09
1.00E-08
1.00E-07
1.00E-06
0.1 1 10
ΔK , MPa m1/2
da/d
n, m
/cyc
le
Δσ = 0Δσ = 40Δσ = 64Δσ = 80Δσ = 100Δσ = 125Δσ = 160Δσ = 200
48
ddaa/d/dnn in AA6082in AA6082--T6 vs. (a) T6 vs. (a) ΔΔKK and (b) and (b) ΔΔKKeqeq
49
ddaa/d/dnn in A508 steel vs. (a) in A508 steel vs. (a) ΔΔKK and (b) and (b) ΔΔKKeqeq
50
Prediction of selfPrediction of self--arresting cracksarresting cracksat notches at notches –– Reanalysis of FrostReanalysis of Frost’’s s
classical fatigue testsclassical fatigue tests
51
Notched fatigue test specimens forNotched fatigue test specimens for(a) push(a) push--pull and (b) rotating bendingpull and (b) rotating bending
52
Stress intensity range corrected with Stress intensity range corrected with respect to (short) crack lengthrespect to (short) crack length
53
Stress intensity range corrected with Stress intensity range corrected with respect to (short) crack lengthrespect to (short) crack length
54
Application of PApplication of P••FAT to the analysis FAT to the analysis of fatigueof fatigue--crack growth from a small crack growth from a small
weld defectweld defect
55
TT--joint made by welding together two joint made by welding together two AA6082AA6082--T6 RHS profilesT6 RHS profiles
56
WWööhler diagram for Thler diagram for T--joint under 4PB: Measured joint under 4PB: Measured and predicted lives incl. the influence of (simulated) and predicted lives incl. the influence of (simulated) residual stresses from welding residual stresses from welding
57
MODELLING THE GROWTH OFMODELLING THE GROWTH OFFATIGUE CRACKS FROMFATIGUE CRACKS FROM
RANDOM DEFECTSRANDOM DEFECTS
58
Different approaches to fatigue analysisDifferent approaches to fatigue analysis
Random DefectSingle DefectExplicit FCG analysisda/dn = f(Δσ, a; R)
Weakest LinkLocal StressImplicit FCG analysisS-N-curve (a > 1 mm)
ProbabilisticDeterministicMaterial properties
Crack Growth
59
P•FAT: Random Defect
For every finite element of a component, random For every finite element of a component, random defects are generated based on the underlying defects are generated based on the underlying statistical distributions (number , location, size).statistical distributions (number , location, size).The defects are The defects are ‘‘rankedranked’’ according to their effective according to their effective stress intensity ranges.stress intensity ranges.This procedure is repeated for a large number of This procedure is repeated for a large number of nominally identical components.nominally identical components.For each component, the number of cycles to failure is For each component, the number of cycles to failure is determined by means of the determined by means of the ‘‘Single defectSingle defect’’ option. option. This is a This is a ‘‘Monte CarloMonte Carlo’’ simulation of the life simulation of the life distribution of the component. distribution of the component.
60
(a) Potentially life(a) Potentially life--controlling defects in controlling defects in a single componenta single component
(b) A priori vs. a (b) A priori vs. a posteriori posteriori ‘‘lifetime lifetime
rankingranking’’
61
LifeLife--controlling defects of 500 nominally controlling defects of 500 nominally equal specimensequal specimens
62
3D distribution of potentially critical defects3D distribution of potentially critical defects
0
100
200
300
400
500
−50
0
50
−100
−50
0
50
100
amin
= 10 μm, amax
= 600 μm, amean
= 40 μm
Truncated exponential distribution
63
Fatigue lives of 1000 nominally equal Fatigue lives of 1000 nominally equal specimens specimens –– Monte Carlo simulationMonte Carlo simulation
1 100 200 300 400 500 600 700 800 900 10000
0.2
0.4
0.6
0.8
1
1.2
1.4
Specimen number
N/N
mea
n
amin
= 10 μm, amax
= 600 μm, amean
= 40 μm
64
Specimens considered to illustrate a Monte Carlo lifeSpecimens considered to illustrate a Monte Carlo life--time simulation based on FCG from random defectstime simulation based on FCG from random defects
65
Probability of failure under homogeneous stress Probability of failure under homogeneous stress based on the Kbased on the K--T diagram and the size T diagram and the size
distribution of defects distribution of defects
66
Monte Carlo simulation of the size of the lifeMonte Carlo simulation of the size of the life--controlling defect assuming the defect size to be controlling defect assuming the defect size to be
GumbelGumbel distributeddistributed
67
Monte Carlo simulation of the smooth and Monte Carlo simulation of the smooth and notched fatigue limits assuming the defect size to notched fatigue limits assuming the defect size to
be be GumbelGumbel distributeddistributed
68
The smooth and notched fatigue limits are well The smooth and notched fatigue limits are well described by the threedescribed by the three--parameter Weibull parameter Weibull distribution:distribution:
This observation connects the This observation connects the randomrandom--defect defect simulation with the simulation with the weakestweakest--linklink approach.approach.
69
Welded pipe subjected to axial tensionWelded pipe subjected to axial tension
70
FE mesh of welded pipe (60,000 elements)FE mesh of welded pipe (60,000 elements)
71
FCG through pipe weldFCG through pipe weld
72
PublicationsPublications
A. Fjeldstad, G. Härkegård, A. WormsenThe influence of a stress gradient on the growth of a fatigue crack. 9th International Fatigue Congress, Atlanta, Georgia, 2006.A. Wormsen, A. Fjeldstad, G. HärkegårdA post-processor for fatigue crack growth analysis based on a finite element stress field. To be published in Computer Methods in Applied Mechanics and Engineering.A. Fjeldstad, A. Wormsen, G. HärkegårdSimulation of fatigue crack growth in components with random defects. To be published in Engineering Fracture Mechanics.A. Fjeldstad, A. Wormsen, G. HärkegårdA reanalysis of Frost’s classical fatigue tests on self-arresting cracks at notches. Dept. of Engineering Design and Materials, NTNU, Trondheim, 2007.
73
ConclusionsConclusions
Fatigue is basically caused by crack growth from stochastically distributed material defectsProbabilistic fatigue assessment may be carried out by means of the analysis of fatigue crack growth from random defectsA tool for such assessment, P•FAT, post-processing stresses from a FEA, is being developed at NTNUP•FAT offers a physically sound and robust method for the fatigue assessment of materials and structures
74
OUTLOOKOUTLOOK
Physically sound and robust models (and data!) for the growth of short fatigue cracksDitto for the initiation of cracks at defects that are not crack-like, e.g., poresStatistical distribution of material defectsVerification testing of real componentsThese are objectives of a Norwegian-Danish program (3.5 MCHF, 2007-2010) on cast components for large wind turbines
75
Seminar presentationsSeminar presentations
This presentation and that from 18.10.2007, ‘A non-local fatigue assessment method based on weakest-link theory and statistics of extremesand its application to component-like specimens’,can both be found as PDFs by means of the linkhttp://www.zfm.ethz.ch/e/v/sem/.