In recent years, much attention has been given to issues of aging,reliability, and the remaining lifespan of the world’s major in-frastructure. As a result of years of neglect, many bridges within theUnited States have been determined to be structurally deficient. In1997, a survey conducted by the New York City Department ofTransportation found that nearly half of the bridges in New YorkCity were in poor condition (the average rating was 4.5, on a scale of1 to 7 with a ranking of 1 considered to be excellent) (Betti et al.
2005). Many of the suspension bridges in the greater NewYork areaalone are approaching or have exceeded 100 years of service life, andin-depth inspections of the cable systems of these bridges haverevealed the presence of many broken wires, showing brittle frac-tures and extensive corrosion (Betti and Yanev 1999).
The life expectancy of a suspension bridge is directly correlated tothe condition of its cable system (formed with thousands of parallelhigh-strength steel wires) and, consequently, to the corrosion of thehigh-strength steel wires (Shi et al. 2007). The corrosion of metalsmay be defined as an electrochemical reaction in which the corrodingmetal is oxidized in the presence of an aqueous environment and anelectrolyte (Jones 1996). Environmental factors such as temperature,moisture, and pH levels are considered agents that may accelerate thechemical processes,which lead togeneral corrosion, pitting, and stresscorrosion, corrosion-induced cracking, and hydrogen embrittlementof high-strength steel wires (Stahl andGagnon 1996; Betti et al. 2005;Cao et al. 2003; Barton et al. 2000; Haynes 1995).
In a given day, bridge cables may be exposed to a variety ofweather patterns (e.g., rain and direct sun), temperature (e.g., dayand night), and moisture levels (e.g., dry and wet). More signifi-cantly, moist air and water may enter the interior of a cable throughany number of openings along the cable (i.e., imperfections of thecoating system and poor compaction of the wires). Once inside,evaporation becomes nearly impossible and the wires may be di-rectly exposed to wetness for extended periods of time (Betti et al.2005). Hopwood and Haven (1984) reported that direct and ex-tended exposure of the high-strength steel wires to moisture in suchenvironments causes a corrosion reaction that is more aggressivethan that created by direct immersion of steel into a corrosive so-lution. The increase of the corrosion product on the surface of the
1Ph.D. Candidate, Dept. of Civil Engineering and Engineering Mechan-ics, Columbia Univ., 638A S.W. Mudd Building, New York, NY 10027(corresponding author). E-mail: [email protected]
2Professor and Chairperson, Dept. of Civil Engineering and EngineeringMechanics, Columbia Univ., 640 S.W. Mudd Building, New York, NY10027. E-mail: [email protected]
steel wires promotes condensation, thus increasing the time ofwetness of the material and, consequently, enhancing the effects ofcorrosion (Roberge 2008). Furuya et al. (2000) found water withinthe main cables of several of Japan’s newest bridges. This waterwas reported to have led to highly humid environments in the maincables resulting in high-strength steel wire corrosion occurring just10 years after bridge completion (Suzumura and Nakamura 2004).
Specific solutes and chemical compounds also enhance theaggressiveness of suspension bridge cables’ environment. An ex-ample is acid rain (with a pH , 5.0); i.e., hydrogen evolution atlow pH levels may cause embrittlement resulting in subsequentcrack initiation/enhancement and material property degradation(Sanchez-Galvez and Elices 1984). Furthermore, Hartt et al. (1993)found that chloride ions from bridge deck deicing salts may pene-trate into the cable, leading to the breakdownof any passivefilms andenhancing the solubility of the corrosion product of carbon steel(found in high-strength steel wires), thus promoting corrosion pit-ting. Eiselstein and Caligiuri (1988) found water with a pH of 4inside themain cables of theWilliamsburg Bridge and estimated thatcable wires were damp for nearly 135 days of the year.Wire samplestaken from that bridge showed pitting corrosion, with a depth of thecorrosion pit that reached up to a 30% loss of diameter.
In this paper, a sensor network, designed to monitor the envi-ronmental conditions [e.g., temperature and relative humidity (RH)]and corrosion activity within a suspension bridge’s main cable, isproposed. The proposed sensing system combines temperature, RH,and corrosion rate sensors to provide both an indirect and directmethod to obtain information regarding the corrosive nature of theinternal environment of a main cable. For a brief analysis of theshortfalls of current inspection techniques, the reader is referred toAppendix S1.A full-scalemock-up cable and environmental chamberwere built to test themonitoring system’s ability to accuratelymeasuretemperature, RH, and corrosion rate. Both graphical analyses andstatisticalmethodswere used to compare the experimental testfindingsto those available in the literature.
Experimental Program
Sensing Technology
Sensors were chosen after an extensive literature review, whichsought sensors that satisfied criteria related to size, accuracy, resis-tance to compaction forces, environmental durability, and sensitivityto environmental variables. The HS-2000V Precon sensor was cho-sen to measure temperature and RH levels within the cable. The HS-2000V provides both temperature and RH readings accurate to6 2%in an operating temperature range of 0–70�C (32–158�F) and ina humidity range of 0–100% (Kele Precision Manufacturing 2011).
The recordings and analyses of two different corrosion ratesensors are presented in this paper. The corrosion rate is the rate atwhich the electrochemical reactions occur on a metal; thus, thecorrosion rate sensor functionality is predicated on the presence ofan aqueous environment formed by atmospheric RH trapped insidethe cable voids—if the RH levels are at 0%, then for each of thesensors considered the corrosion rate recordings are zero. The firstcorrosion rate sensor considered is the Analatom (2011) linearpolarization resistance (LPR) sensor, which uses the linear polari-zation technique to estimate corrosion rates. Linear polarization is anelectrochemical technique in which a potential scan is applied toa freely corroding sensor element and the resulting current is mea-sured.AnalatomLPR sensors use two electrodes, a working electrodeand a counter/reference electrode, which aremade from shim stock ofAISI 1080 steel (AISC 2011) (a similar chemical composition to that
of the high-strength steel wire used in this experiment) (Analatom2011). The second corrosion rate sensor considered in this study is theCorr Instruments (2011) coupled multiple array sensor (CMAS),which consists of miniature electrodes made of the corrodingmaterialof interest. Based on metal composition, some electrodes have moreanodic properties while others have cathodic properties. Whencovered by a corrosive aqueous solution, CMASs electrically connectanodic and cathodic sites of the same metal. The resulting electronflowbetween the aforementioned sites ismeasured and then convertedinto nonuniform corrosion rates using Faraday’s law (Corr Instru-ments 2011). Other types of sensors were considered in the study;however, the analysis of such sensors is not presented in this paper.
Cable Specimen
To test the effectiveness of the monitoring system, a full-size mock-up cable specimen was built and exposed to varying controlledenvironmental conditions. The main cables of suspension bridgesare made of thousands of compacted, parallel high-strength steelwires, spun either singly or in prefabricated strands, with a void ratioof roughly 20%. The mock-up specimen, a 1:1 scale replica ofa main cable of a long suspension bridge, was made with 73 hex-agonally shaped strands, each consisting of 127 high-strength steelwires. Each strand was fabricated by placing the 127 wires ina special comb, custom-built in the laboratory, and then compactingthem using specially constructed hexagonal aluminum clamps. Thecore of the cable was composed of 61 of the 73 strands; the re-maining 12 were used to fill voids and for the creation of a circularcross section following cable compaction. Thus, a cable was createdwith a 50.8-cm diameter and a cross-sectional area composed of9,271 wires. Of the 61 core strands, 54 measured 6.10 m in length,a length that is on the order ofmagnitude of the distance between twoadjacent vertical suspenders. The remaining seven strands, 10.67-mlong, were subjected to a total tensile load of over 4,448.22 kN inorder to induce stresses in the wires up to 700 MPa (a stress levelslightly higher than that experienced in service load conditions).This stress was used to highlight and eventually accelerate thephenomenon of stress-corrosion cracking.
Individual steel wires, with a diameter of 0.498 cm, were gal-vanically protected by Class A zinc coating. The nominal ultimatetensile stress of the wires was 1,700 MPa, the yield stress was ap-proximately 1,400 MPa, and the elastic modulus was 200,000 MPa.The chemical composition of the galvanized wires was as follows:0.80–0.81% carbon, 0.81–0.82% manganese, 0.23–0.27% silicon,0.07–0.08% chromium, 0.06% nickel, 0.006% sulfur, and 0.003%phosphorous, and the remaining composite was iron. The compo-sitions were within the typical ranges, as provided in Stahl andGagnon (1996) for high-strength bridge wires.
Following construction completion, the cablemock-up specimenwas wrapped with aluminum foil tape. The tape served doubly asa wrapping and encasement for the high-strength steel wires, whichprotected them from direct contact with the environmental cham-ber’s atmospheric conditions while ensuring that high levels of RHand temperature surrounded thewires. In fact, while in real cables thepurpose of the wrapping is to protect the wires inside, in this study itwas exactly the opposite; i.e., the purpose was to create increasedlevels of temperature and RH in order to maintain a highly corrosiveenvironment in which to test the sensor system.
Sensor Arrangement
A total of 50 sensors were installed throughout the cross section ofthe mock-up cable specimen; however, only the results from 29 ofthe 50 sensors are presented in this paper (Fig. 1). To obtain a spatial
distribution of the variation of the various variables of interest(temperature, RH, and corrosion rate), the 50 sensors were placed atvarious depths (outer, middle, and inner locations within the cable)along three directions (or diameters) inclined at 60� with respect toone another. The evenly spaced placement of the sensors provided anexperimental image of the distributions of temperature, RH, andcorrosion activity within the cable. To protect the sensors fromcrushing during cable compaction, they were placed in between two2.5-cm-long stainless steel pipes covered with a heat-shrinking,moisture-resistant coating (in order to prevent the galvanic corrosionof the steel wires). The sensors were hardwired and the electrical wireswere run out of the cable center at two vertical locations and connectedto a data acquisition system called Sensor Highway II, developed byMistras Group, Inc. (Mistras Group, Inc. 2012). Fig. 2 shows the in-stallation of both the Precon HS-2000V and Analatom’s LPR sensors.
The sensor numbering displayed corresponds to the numberingused during data collection. Seventeen PreconHS-2000V temperature/RH sensors (labeled T in Fig. 1 but referred to as T or RH dependingon whether temperature or RH recordings are in question) wereevenly distributed according to the depth and angle of incline.Sensor T6 was discarded from use (not referenced in Fig. 1) be-cause of early installation malfunctions. Both temperature and RHdata were collected at 300-s intervals for the duration of each test.
Eight Analatom LPR sensors (labeled LP in Fig. 1 but referred toas LPR) were distributed in the upper half of the mock-up cablespecimen’s cross section. Four sensors were placed along the uppervertical radius, while twowere placed on each of the radii oriented at6 60� from the vertical location (Fig. 2). A complete cross-sectionaldistribution of the corrosion rate sensors was desirable; however,laboratory construction operation forced the placement of the LPRsensors to the upper half of the cable’s cross section. The LPRsensors collected corrosion rate data every 60 s.
Four CMASs [one carbon steel (CS) sensor, CMAS CS, and onezinc sensor, CMAS Zn)] were located at the top of the cable (twosensors) and at the bottom of the cable (two sensors). All recordedthe corrosion rate every 300 s. Prior to large-scale testing, theCMASs were tested and calibrated by the manufacturer while theLPR and Precon sensors were extensively tested in accordanceto ASTM Specification G85-02 (ASTM 1994) (and calibrated fol-lowing testing) in a 1,100-L capacity Q-FogCyclic Corrosion Tester
available at the Carleton Laboratory at Columbia University. Theremaining 21 sensors (not displayed in Fig. 1) measured the pH andcorrosion rate levels (using various methods); however, they are notincluded in this paper.
Fig. 1. Schematic drawing of a cable mock-up cross section with embedded sensors
Fig. 2. Placement, protection, and hard wiring of Precon HS2000V(dark, rectangular sensor head on extreme left of image) and AnalatomLPR sensors (covered by light mesh and adjacent to Precon HS2000Vsensor)
For a brief description of the cyclic corrosion testing chamber, thereader is referred to Appendix S2. The cable mock-up specimenwas exposed to a variety of cyclic environmental conditions,consisting of various combinations of rain, heat, air conditioning,and ambient conditions. In this paper, the results from the fol-lowing test dates are presented: Test 1 (May 23, 2009), Test 2(August 10, 2009), Test 3 (August 26, 2009), and Test 4 (Sep-tember 15, 2009). The composition of such tests is presentedin Table 1.
Experimental Results/Discussions
Temperature/Relative Humidity Analysis Overview
In this section, the general temperature and RH trends are discussedusing the measurements from Test 1. This test was chosen because(1) a Precon HS-2000V sensor (labeled AT2) was placed inside thechamber in order to relate the chamber environment to the envi-ronment within the cable specimen; (2) from the results, the RHlevels within the cable varied significantly, allowing for a betterunderstanding of the RH distributions across the cable’s cross sec-tion; and (3) the initial temperature levels were recorded at ap-proximately room temperature throughout the cross section, makingit possible to clearly highlight the effects that the proximity to theheat source and the cable surface had on the temperature distributioninside the cable specimen.
In the data analysis, the initial data recordings were smoothedusing a local regression scheme with a weighted linear least-squares and a second-degree polynomial model in which the out-lier data points received a lower weighting than the others. Oncesmoothed, cyclic maxima and minima were identified and thetemperature gradients were found for various locations withinthe cable cross section. The gradients may be compared with thehighlighted effects of the distance from the heat source and ofthe atmospheric conditions on the temperature variation within thecable.
The analysis of the temperature recordings consists of two parts.The first part considers the chamber temperature variations while thesecond part considers the distributions of temperature in the uppersection of the cable (similar conclusions can be drawn for the lowerportion of the cable). Because of significant differences in thetemperature values and variations, the analysis of the upper sectionwas subdivided based on three levels (inner, middle, and outer) ofthe sensor distance from the center of the cable’s cross section. In theupper portion of the cable cross section, the inner level consisted ofSensors T12, T4, and T13, while the middle and lower levelsconsisted of Sensors T11 and T14 and Sensors T5 and T15, re-spectively (the temperature component of Precon Sensor 10 did notfunction). For the lower portion of the cable’s cross section, the
inner, middle, and outer levels consisted of Sensors T9, T16, and T2,Sensors T8 and T17, and Sensor T1, respectively (Sensors T7 andT18 did not function following installation).
Cyclic Chamber Temperature Recordings (AT2)
In Test 1, the minimum temperatures recorded within the chamberby Precon Sensor AT2 ranged from approximately 67.4 to 73.9�F(Fig. 3). These minima increased on average 1.3�F per cycle andwere recorded at the end of the rain phase for each cycle; i.e., 0.5 hfollowing the start of a cycle. The onset of the heat phase of eachcycle drove the temperature values to their maxima (ranging from115.6 to 119.5�F); i.e., the levels that were reached at the end ofeach heat phase (1 h after the minimum values were recorded). Fol-lowing the temperature increases, the heat and air-conditioningphase of testing created a slight drop in temperature values of ap-proximately 7–9�F less than each cycles’ respective peak value.Temperatures fluctuated about these values for the remainder ofeach of the 1-h long heat and air-conditioning phases and thendropped over the final 1.5 h to levels 2–3�F higher than the pre-vious cycles’ minimum value. In general, only minor increases ofchamber temperature minima were recorded with each cycle in thetest’s progression.
Upper Cable Section Temperature Recordings
Sensor T3, being centrally located within the cable’s cross section,was taken as the baseline and its recordings are shown in Figs. 4(a–c)for comparative purposes; it recorded temperature values between72 and 85�F showing no distinct cyclic minima and maxima. Atemperature increase of 5�F per cycle and a decrease of about 1�F foreach cycle were estimated. Although slight fluctuations appeared,a linear regression analysis provided an r-squared value of 0.9606for the temperature increases recorded by T3 [Fig. 4(a)], indicating
Table 1. Environmental Chamber Cyclic Corrosion Test Descriptions
Cycle phaseCable wrappingcoverage holeTest Total cycle length Rain Heat Heat/air conditioning Air conditioning Ambient Ventilation
1 4 0.5 1 1 1.5 — Open None2 6 1 3 — — 2 Closed Two 30:53 30:5 cm;
two 2:543 2:54 cm3 6 1 3 — — 2 Open Two 30:53 30:5 cm;
two 2:543 2:54 cm4 4 1 1 1 1 — Open Two 30:53 30:5 cm;
two 2:543 2:54 cm
Fig. 3. Test 1: Precon HS2000VAT2 chamber temperature recordings
that a linear increase of the temperature at the center of the cablewith time is a valid assumption for such a cyclic test becausethe temperature appears to monotonically increase with a rate of1.07�F/h.
As expected, the greatest temperature increases and fluctuationswere recorded in the outer level of the cable’s upper section becauseof the proximity of the sensors to the heat sources. The outer level[T5 and T15, Figs. 4(b and c)] recorded average temperature minimain the range of 72–92�F at approximately 5–10 min following thecompletion of each cycle’s rain phase while average maxima,recorded approximately 2.5 h after minimum recordings for eachcycle, gradually increased from 84 to 102�F. Cyclic temperatureincreases of approximately 12.5�F per cycle were recorded, withcorresponding decreases of approximately 7.5�F per cycle. Slightdifferences in temperature gradient, within the measurement noise,existed between the recordings of T5 and T15, with average cyclicincreases of 5.16�F/h (T5) and 4.31�F/h (T15). In the outer sensorlevel, moving from the center to the edges, the temperatures de-creased at a rate of 0.014�F/h/degree (from the vertical location).
Cyclic temperature gains and fluctuations were also recorded bysensors in the middle level segment [T11 and T14, Figs. 4(a and c)].The temperature minima ranged from approximately 71 to 87.5�Fover the entire course of testing, with the lowest temperaturesreached at approximately 10–15 min following the minimum valuesof the chamber’s temperature. In each cycle, temperature increases,averaging 8.5�F per cycle, occurred over the 2.5-h period followingtheminimum recordings, with averagemaximumvalues in the rangeof 78.5–94�F. For the heating phase of each cycle, the temperaturegradients on either side of the cable at the midlevel were approxi-mately 2.350�F/h (T11) and 2.8�F/h (T14), suggesting that the lefthalf and the right half of the cable heated similarly. Cyclic tem-perature decreases were steadier in the middle level compared withthe outer level, averaging 5�F per cycle.
As expected, the inner level sensors [T12, T4 and T13,Figs. 4(a–c)] recorded the slightest fluctuations and smallest tem-perature increases in the cable’s upper portion, with temperatureminima and maxima ranging between 71–89�F and 79–94�F, re-spectively, with an average 7�F increase and 3�F decrease in tem-perature per cycle. During the heating phase of each cycle, T12 andT13 recorded temperature increase rates of 1.740 and 1.662�F/h,respectively, while T4 recorded a slightly higher value (1.8479�F/h).This was expected because T4 was in closer proximity to theheat source. Along the inner level, the temperature gradients de-creased by an average of 0.0024�F/h/degree (from the verticallocation), validating that even at the inner level both halves of thecable (left and right) heated evenly, with the center portion (directlyunder the heat source) heating at a slightly greater rate than thesides.
Fig. 4. Test 1: upper cable temperature distributions from vertical locations along the diameters at (a) 260�, (b) 0�, and (c) 60�
Fig. 5. Test 1: Precon HS2000V ARH2 chamber RH recordings
At the inner level, temperature minima were recorded approxi-mately 20min after those recorded in the chamber. The time intervalbetween the minimum and maximum temperature at the inner levelwas the same as those recorded at both the middle and outer levels;however, therewas a time lag of 10–15 min between these minimumand maximum temperatures and the corresponding temperatures atthe other outer levels. This lag was a result of the thermal diffusivityproperties of the cable. Furthermore, because of their proximity tothe heat source, Sensors T5 and T4 recorded maximum temperaturevalues that were, on average, 1–2�F higher than those recorded bythe other sensors in their respective levels.
The results of Test 1 show that with greater distance from the heatsource the temperature variations within the cable’s cross sectiondiminished with respect to the outside temperature. The maximumtemperatures within the cable did not reach levels as high as thoserecorded in the chamber, and the temperature fluctuations decreasedwith increased distance from the heat source. The upper outer regionshowed substantial temperature fluctuations, whereas near mono-tonic increases in temperature occurred in the center of the cable. Theaverage temperature gradients found during the heating phase ofeach cycle prove that the cable heated evenlywithmaximumheatingrates being obtained at central locations. Finally, the time to whichthe cable interior was affected by temperature fluctuations increasedwith greater cable depth.
While the temperature data were used in collective analysis, generaltrends were not as identifiable in the RH data. Despite incon-sistencies, the RH data are also presented based on the aforemen-tioned chamber location and various levels of sensors within the
upper section of the cable. The recordings of Precon Sensor ARH2once again displayed the effects of the various phases of the cycle onthe RH levels within the testing chamber. The onset of the rain phasein Test 1 caused the chamber RH to increase from 55 to 99% (Fig. 5).Approximate 15% decreases were recorded in the peak sections ofthe first two cycles of RH. These drops may be considered noisebecause the remainder of testing displayed stable peak RH readings.Sensor ARH2 recorded peak RH values for a time extending, onaverage, approximately 10 min longer than the length of the rainphase. Over the course of approximately 1 h, corresponding to thelength of the heat phase, the RH levels within the chamber droppedto an average value of 16%. The subsequent cyclic shifts to 1-h-longphases of heat and air conditioning resulted in plateaued levels of thechamberRHat 20%.Thefinal 1.5 h of air conditioning for each cycleled to chamber RH increases to levels of 55%. The final cycle of Test1 was not completed, and thus the RH percentages did not reach thepreviously discussed maxima (Fig. 5).
Upper Cable Section Relative Humidity Recordings
As in the temperature analysis, the RH recordings of Precon Sensor 3(RH3), referring to the center of the cable, are used for comparativepurposes. The sensor recorded initial RH levels of 40% while thereading within the chamber was at 55%. Over the course of testing,despite the external fluctuations in RH levels, the center of themock-up cable experienced an approximate cyclic decline in RH of 2%. Alinear regression analysis returned an r-squared value of 0.9701,suggesting a nearly linear decline of the RH in the cable center. Thiswas perfectly consistent with the linear increase of the temperatureat the center, as shown previously.
Fig. 6. Test 1: upper cable RH distributions from vertical locations along the diameters at (a) 260�, (b) 0�, and (c) 60�
Sensors within the outer level (RH10, RH5, and RH15) of theupper portion of the cable mock-up specimen recorded substantialvariations in RH levels (Fig. 6). Sensor RH10 recorded the highestinitial RH levels at 88% [Fig. 6(a)]. This may be a result of one of thefollowing causes: (1) some local trapped water; (2) a sensor near tocomplete saturation; and (3) sensor malfunction [the temperaturecomponent of Precon Sensor 10 (T10) was not functioning, thus theRH readings may also be inaccurate].
Sensor RH5, located on the upper section of the cable, recordedpeak values corresponding to the onset of the rain phase [Fig. 6(b)].At the beginning, RH was maintained at levels of approximately43% for 1 h, eventually declining to 31%. The RH peaks declinedfrom 37 to 33%, at an average rate of about 1% per cycle. For 2.5 hfollowing each cyclic maximum, the RH percentages experiencedaverage per cycle decreases of approximately 6%. The RH valueswere found to increase, on average, 5% over the final 1.0 h of eachcycle—an increase in RH that corresponded, as expected, to atemperature decrease. The levels continued to increase for 0.5 h withthe onset of each new cycle. Because of the proximity of RH5 to theheat source, the subsequent heat phases seemed to have dried thesensor, thus driving the RH values downward.
Sensor RH15 recorded initial levels of RH at 68% (higher thanthose recorded in the chamber). However, unlike ARH2, RH15[Fig. 6(c)] recorded RH increases of approximately 4% at a time 1.5h after the onset of each cycle’s rain phase. The RH levels fluctuatedabout these points with slight decreases for the next 2.5 h of eachcycle. Sharp declines of approximately 5% were recorded at a timecoinciding with the start of the subsequent cycle’s rain phase,corresponding to cooling of the cable. The overall decline of RHshown by RH15 was of approximately 10% over the course oftesting.
Unlike those in the outer level, the middle level sensors [RH11and RH14, Figs. 6(a and c)] showed similar trends to one another.RH11 recorded initial levels of RH at approximately 57%. Onceagain, at a time 1.5 h following the onset of cyclic rain phases, RH11recorded RH increases of approximately 7%, keeping this levelalmost constant for 2.5 h for each cycle. At the initiation of sub-sequent rain phases, RH11 recorded average RH drops of 10%, witha final RH level of 50%. A similar pattern was recorded by RH14[Fig. 6(c)], with an initial RH reading of 58% and a final readingof 52%.
Still in the middle level, RH4 recorded fluctuations oppositethose of RH5 [Fig. 6(b)]. Starting from an initial level of 42%, at theonset of each rain phase, RH4 showed an average decrease in RH ofapproximately 4% occurring over the course of 50min to 1 h. Cyclicincreases of approximately 3%were recorded over the following 1.5h, followed by a plateau for 2 h prior to declining at the onset of thesubsequent cycles rain phase. RH values eventually reached ap-proximately 26%.
With regard to the inner sensor, those located on the diagonals inthe upper section of the cable’s cross section (RH12 and RH13)recorded near steady declines over the course of Test 1. This agreeswith the fact that the internal temperature of the cable steadilyincreases. RH12 [Fig. 6(a)] recorded initial RH levels of 53% thatslowly decreased to a final value of 40%. Similarly, RH13 [Fig. 6(c)]showed a decline from approximately 43–30%, closely resemblingthe recordings of RH3.
In general, the RH levels recorded during Test 1 varied signifi-cantly across themock-up cable cross section. Sensors located closerto the cable surface (e.g., RH15 and RH10) recorded elevated RHlevels while those closer to the cable center recorded substantiallylower levels. Additionally, close to the cable’s core, sensors recor-ded near-linear monotonic decreases in RH of approximately 2%per cycle corresponding to nearly linear increases in temperature of
approximately 5�F. Conversely, over the course of testing thosesensors closer to the surface displayed greater cyclic fluctuations.
Although the increased levels of RH seemed to correspond wellwith increased distances from the cable core (i.e., closer proximity tothe cable surface), the levels of cyclic RH fluctuations did not. Thiscould be a result of the fact that water penetrated through variouspaths into various parts of the cross section—especially in the outerlayers. In general, the RH sensors recorded increased levels ata period of 1.5 h after the onset of cyclic rain phases. Significantshifts in the time at which increased and decreased levels of RHwere recorded were not observed with shifts in the levels of sensordepth.
The role of temperature and sensor saturation must be consideredas areas in which the cable may be prone to collect or expungewater,thus resulting in inexplicable levels and fluctuations of RH. Manysensors recorded prolonged periods of elevated RH, suggesting thatwater introduced into the cable during the rain phase could notevaporate during the cyclic heat phases. The entrapment of moisturewithin the mock-up cable represents a very serious problem in thedeterioration of suspension bridge main cables. Specifically, theinability for evaporation to occur results in an aggressive environ-ment in which corrosion may occur at elevated rates and sensingtechnology may be negatively affected.
This conclusion is further bolstered by consideration of the RH5sensor recordings. The RH levels recorded by RH5 peaked with theonset of the rain phase. The peak levels proceeded to decline duringthe heat phase, eventually increasing with the onset of air condi-tioning. The lowering of RH with increased temperature wasexpected because increasing temperature promotes evaporation ofwater molecules from an electrode surface. The position of PreconSensor 5 at the top of the cable and its proximity to the cable surfacemay allow for the excess water/moisture to both drain and/orevaporate from the sensor surface. Furthermore, condensation andan increase of RH with decreasing temperature were both expectedand seen in the recordings of Precon Sensor 5 during Test 1. Thedistribution of the RH, recorded by the sensors over the cross section,further validated the 2004 findings of Suzumura and Nakamura(2004), in which decreased levels of RH were identified in theupper portion while higher levels of RHwere found in the sides andlower portion of the cross section.
The temperature and RH recordings from this test confirm theexpectation of an inversely proportional relationship between thesetwo environmental variables (Bohren and Albrecht 1998; Lawrence2005). A statistical analysis of the temperature and RH levelsrecorded by Precon Sensors 5, 3, 13, and 12 returned correlationcoefficient values of less than 20.95, showing an extremely stronginversely linear relationship between the respective temperature andRH recordings. Unfortunately, because of the potential presenceof pockets of water, moisture, and sensor saturation, not all corre-lations were as strong. Thus, the RH levels cannot be strictlyexplained by temperature recordings, instead they must be con-sidered individually.
Corrosion Rate Analysis: Linear PolarizationResistance Sensors
Of the eight LPR sensors placed in the cable, only LPR3 and LPR4recorded corrosion rates above the noise values. The analysis of thecorrosion rate sensors was twofold. The corrosion rate recordingsfrom each sensor were compared and contrasted with the localtemperature and RH recordings, as well as other corrosion ratesensor recordings. The effect of both the environmental conditionsand the depth (location) within the cable cross section on corrosionwas addressed by considering the correlation between atmospheric
variables and corrosion rate fluctuation. For a better understandingof the interaction between temperature/RH and corrosion rate, therecordings of LPR3 and LPR4 are presented against T13 and T15,respectively, for Tests 1 and 3. These tests were selected because,even though the temperature ranges were similar, the RH valuesdiffered enough between tests to permit an analysis of the effectsof both environmental variables. The corrosion rate values recordedby the LPR sensors appeared to follow distinct cyclic trends asso-ciated with temperature, while the initial rate levels were affectedby the RH.
Linear Polarization Resistance Corrosion RateRecordings: Test 1
On May 23, 2009, LPR3 recorded initial corrosion rates at ap-proximately 62 mm/year. Corrosion rate minima were recordedbetween 62 and 73mm/year and the cyclicmaxima ranged from65 to74 mm/year [Fig. 7(a)]. The corrosion rate values began to increaseapproximately 1.75 h after the onset of the test’s rain phase, with anincrease of 3 mm/year over the span of approximately 2.5 h for eachcycle, and remained stable at the increased levels for periods rangingfrom 1.75 to 2 h. This period of plateaued levels existed for eachcycle.
Concurrently, although the RH values recorded by RH13steadily declined from 43 to 30% [Fig. 7(a)], the temperature
increased both cyclically and over the entire test duration. Tem-perature minima [Fig. 7(b)] were recorded approximately 1 h and 10min after the onset of each cycle’s rain phase, with values rangingfrom 71–89�F. Maximum temperatures were recorded as approxi-mately 79–94�F, with temperature variations that showed averageincreases of 7�F and average decreases of 3�F per cycle. Temper-ature increases of 7�F per cycle corresponded to corrosion rateincreases of 3 mm/year, while temperature decreases of 3�F cor-responded to negligible corrosion rate decreases.
A statistical analysis revealed that in Test 1 the LPR3 recordingshad strong linear and negatively linear relationships with tempera-ture and RH levels, respectively. A correlation coefficient of 0.9855existed between the data recorded for LPR3 and T13, while a cor-relation coefficient of20.9844 existed between those recordings ofLPR3 and RH13. The statistical data suggest that the cyclic trends incorrosion rate readings by the LPR sensors were directly related tothe temperature increases.
Closer to the cable surface, the corrosion rate, RH, and temper-ature fluctuation increased substantially. As shown in Fig. 8, LPR4recorded initial corrosion rates of approximately 86 mm/year, witha RH of 68% [Fig. 8(a)]. At approximately 45 min after the onset ofeach cycle’s rain phase, LPR4 recorded corrosion rate average in-creases of 8 mm/year over a period of approximately 2 h, followedby a staged decrease of approximately 5mm/year (mild decreases ofabout 2mm/year over a 1-h period followed by drops of 3mm/year in
Fig. 7. Test 1: (a) LPR 3 versus RH13; (b) LPR 3 versus T13
Fig. 8. Test 1: (a) LPR 4 versus RH15; (b) LPR 4 versus T15
the following 30–45 min). The corrosion rate minima ranged from86–104 mm/year, while maximum values ranged between 94 and110 mm/year.
During the same time period, T15 recorded average temperatureminima in the range of 72–92�F [Fig. 8(b)]. These values also oc-curred at approximately 45 min following the start of each cycle’srain phase. The average maxima were recorded approximately 2.5 hafter theminimum recordings for each cycle and gradually increasedfrom 84 to 102�F. Slight shifts in slope were seen in the decreasingrecordings of T5 and T15; however, no evidence of a definitivetemperature plateau existed (as seen in the chamber recordings).
During Test 1, a correlation coefficient of 20.3192 existed be-tween the LPR4 and RH15 RH recordings, while with the T15temperatures the correlation coefficient was 0.9841; thus confirmingthe strong linear relationship between corrosion rate and temperaturereadings. A temperature increase of 12.5�F per cycle correspondedto a cyclic corrosion rate increase of 8 mm/year and a temperaturedecrease of 7.5�F per cycle corresponded to a cyclic corrosion ratedecrease of 5 mm/year. Thus, for each 61.5�F increase in tem-perature, there was a 61 mm/year increase in corrosion rate. Com-paring themeasurements fromRH13 and RH15, a difference in RHlevels of 25% appears to correspond to an increase in baselinecorrosion rates of 24 mm/year between LPR3 and LPR4. For theoverall test, the temperature increases of 23�F recorded by T13 ledto increases in the corrosion rate recordings by LPR3 of 12 mm/year, whereas the temperature increases recorded by T15 of ap-proximately 30�F resulted in corrosion rate increases of 24 mm/year by LPR4, keeping almost identical increase and decrease rateswithin cycles.
Linear Polarization Resistance Corrosion RateRecordings: Test 3
A statistical analysis of the relationship between the data recordedby LPR3 and T13 and LPR4 and T15 again revealed a strong linearrelationship between the corrosion rate and temperature with cor-relation coefficients of 0.9899 and 0.9928, respectively. The cor-rosion rate recordings from both LPR3 and LPR4 showed anincrease of 1mm/year per 1�F temperature variation (LPR3: 27 mm/year for 25�F overall, 10 mm/year for 9–10�F increase per cycle;LPR4: 28mm/year for 33�F increase overall, with an average 12mm/year for a cyclic average 11–12�F increase per cycle). As seen in Test1, the Test 3 corrosion rate fluctuations did not strongly relate tothe RH fluctuations.
Discussion: Linear Polarization ResistanceCorrosion Rate
From the analysis of the recorded data it appears that RH controls themajor (overall) levels of corrosion rate fluctuations while tempera-ture changes affect themore localized and cyclic corrosion rates. Theeffects of RH were shown in the readings of LPR3 and RH13compared with those of LPR4 and RH15. Similar results may beseen by comparing data from various tests for a specific locationwithin the cable. For example, RH15 recorded a 14% increase in RHlevels betweenMay 23, 2009 (Test 1), andAugust 26, 2009 (Test 3),which corresponds to a 50 mm/year increase in corrosion rates, asrecorded by LPR4.
A statistical analysis revealed that the LPR recorded corrosionrate fluctuations have a strong linear relationship to temperaturereadings. At increased levels of RH (Test 3) a shift in temperature ofapproximately 1�F resulted in a shift of approximately 1 mm/year incorrosion rate. At lower humidity levels, the ratio changes becausea 1�F shift in temperature leads to a shift in corrosion rate that is less
than 1mm/year (Test 1). Such an effectmay be seenwhen comparingthe recordings of LPR3 and T13with respect to LPR4 and T15 in theMay 23, 2009, test [Fig. 8(a) versus Fig. 8(b)]. Thus, depending onthe RH levels and the initial corrosion rates, it may be possible todetermine a proportionality constant between the temperature andcorrosion rate and use the temperature data to predict and verifycorrosion rates within the mock-up cable’s cross section.
Corrosion rate readings from the CMAS CS sensors were analyzedin conjunction with the temperature and RH levels recorded by T5.In a similar fashion to the LPR analysis, the CMAS analysis iden-tified cyclic fluctuations in corrosion rate readings as well as thefluctuations of the surrounding environmental variables. The sen-sors being considered were located at the top of the cable crosssection at 0� from the vertical axis (in Fig. 1, MC2 and T5). In thispaper, the CMAS Zn recordings have been omitted because em-phasis has been placed upon the analysis of the corrosion of ferrousmetal. Because of the various characteristics, test results are pre-sented for two different conditions; i.e., corrosion rates forRHbelow50% and corrosion rates for RH between 80 and 90%. Althoughgeneral trends were identifiable, specific conclusions relating thecorrosion rates and atmospheric variables were difficult to ascertain.For further information regarding the CMAS analysis, the reader isdirected to Appendix S3.
Conclusions
The proposed sensor network system was able to provide an un-derstanding of the interior environment of a suspension bridge’smain cable. The construction of an environmental testing chamberpermitted the testing of various environments affecting the tem-perature, RH, and corrosion rate distributions within the mock-upcable. Compared with the external chamber environment, cyclicenvironmental fluctuations were reflected in outer sensor recordingsof temperature and RH, while sensors in the core of the cablerecorded near constant increases in temperature and decreases inRH.The temperature levels and fluctuations proved to vary according todepth within the cable.
Although they did not follow consistent trends across the cable’scross section, the RH values were strong indicators of corrosion ratelevels, as validated by those recorded by both the Analatom (2011)LPR and Corr Instruments (2011) CMASCS corrosion rate sensors.A corrosion rate analysis highlighted the distinct differences in thedependency of the corrosion process on the environmental variablesof temperature and RH. Increased levels of RH resulted in increasedlevels of corrosion activity recorded by both types of sensors.Furthermore, statistical analyses showed that the experimental de-pendence of corrosion rate values, as recorded by the LPR sensors,on temperature was strongly linear. In addition, while the temper-ature values varied with cable depth and followed distinct trends incyclic fluctuations, the RH did not.
The formation of predictive models for the distribution of bothtemperature and RH within suspension bridge main cables is im-perative. Fluctuations in temperature and RH were recorded andconsidered in order to provide information to bolster the de-velopment of robust mathematical models. Currently, models fortemperature distributions across a cable cross section are beingestablished. The current system is being prepared for installation onan in-service bridge. Field testing will be used to determine the real-world functionality of the system. Future studies will compare the
experimental and in-service sensor recordings for all variables(temperature, RH, and corrosion rate) and will develop methodol-ogies to include sensor data in the estimation of the remainingstrength of the main cables of suspension bridges.
Acknowledgments
This study was sponsored by the Federal Highway Administra-tion under Contract No. DTFH61-04-C-00040 (program managers,Dr. H. Ghasemi and Dr. P. Virmani). The support and guidance ofDr. Ghasemi and Dr. Virmani are greatly appreciated. The continu-ous suggestions by Dr. Bojidar Yanev, New York City Departmentof Transportation, were also greatly valuable. The authors wouldlike also to acknowledge the contribution of Mr. M. Carlos andMr. R. Gostautas from Mistras Corporation for their help withthe data acquisition system. A special thanks is extended to Dr.B. Laskowski, from Analatom Corporation, for his assistance withthe LPR sensors.
Supplemental Data
Appendixes S1–S3 and Figs. S1–S3 are available online in theASCE Library (www.ascelibrary.org).
References
AISC. (2011). Steel construction manual, 14th Ed., American Institute ofSteel Construction, Chicago.
ASTM. (1994). “Standard practice for modified salt spray (fog) testing.”G85-02, West Conshohocken, PA.
Barton, S. C., Vermaas, G.W., Duby, P. F.,West, A. C., andBetti, R. (2000).“Accelerated corrosion and embrittlement of high-strength bridge wire.”J. Mater. Civ. Eng., 12(1), 33–38.
Betti, R., West, A. C., Vermaas, G., and Cao, Y. (2005). “Corrosion andembrittlement in high-strength wires of suspension bridge cables.” J.Bridge Eng., 10(2), 151–162.
Betti, R., and Yanev, B. (1999). “Conditions of suspension bridge cables:The NewYork City case study.” Transportation Research Record 1654,Transportation Research Board, Washington, DC, 105–112.
Bohren, C., and Albrecht, B. (1998). Atmospheric thermodynamics, OxfordUniversity Press, New York.
Cao, Y., Vermaas, G. W., Betti, R., West, A. C., and Duby, P. F. (2003).“Corrosion and degradation of high-strength steel bridge wire.” Cor-rosion, 59(6), 547–554.
Eiselstein, L. E., and Caligiuri, R. D. (1988). “Atmospheric corrosion of thesuspension cables on the Williamsburg Bridge.” Degradation of metalsin the atmosphere, ASTM STP 965, S. W. Dean and T. S. Lee, eds.,ASTM, Philadelphia, 78–95.
Furuya, K., Kitagawa, M., Nakamura, S., and Suzumura, K. (2000).“Corrosion mechanism and new protection methods of suspensionbridge cables.” Struct. Eng. Int., 10(3), 189–193.
Hartt, W. H., Kumria, C. C., and Kessler, R. J. (1993). “Influence ofpotential, chlorides, pH, and precharging time on embrittlement ofcathodically polarized prestressing steel.” Corrosion, 49(5), 377–385.
Hopwood, T., and Haven, J. (1984). “Inspection, prevention, and remedyof suspension bridge cable corrosion problems.” Kentucky Transpor-tation Research Program Rep. U.K.TRP-84-15, Univ. of Kentucky,Lexington, KY.
Jones, D. A. (1996). Principles and prevention of corrosion, 2nd Ed.,Prentice Hall, Upper Saddle River, NJ.
Lawrence, M. G. (2005). “The relationship between relative humidity andthe dewpoint temperature in moist air: A simple conversion andapplications.” Bull. Am. Meteorol. Soc., 86(2), 225–233.
MistrasGroup, Inc. (2012). “Sensor highway II: OutdoorAE and sensor fusionsystem.” Æhttp://www.mistrasgroup.com/products/company/Publications/2$Acoustic_Emission/Sensor_Highway_II.pdfæ (Mar. 27, 2012).
Roberge, P. (2008). Corrosion engineering: Principles and practice,McGraw Hill, New York.
Sanchez-Galvez, V., and Elices,M. (1984). “On hydrogen-induced crackingin prestressing steel wires.” Proc., 5th European Conf. of Fracture(ECFS), EMAS, South Yorkshire, U.K., 1003–1014.
Shi, Y., Deodatis, G., Betti, R. (2007). “Random field-based approach forstrength evaluation of suspension bridge cables.” J. Struct. Eng.,133(12), 1690–1699.
Stahl, F. L., and Gagnon, C. P. (1996).Cable corrosion in bridges and otherstructures: Causes and solutions, ASCE, New York.
Suzumura, K., and Nakamura, S. (2004). “Environmental factors af-fecting corrosion of galvanized steel wires.” J. Mater. Civ. Eng.,16(1), 1–7.
