Damage Identification of a Long-Span Suspension Bridge Using Temperature- Induced Strain Data. Southeast University, Nanjing, China, ABSTRACT

Transcription

1 Damage Identification of a Long-Span Suspension Bridge Using Temperature- Induced Strain Data *Qi. Xia 1) and Jian. Zang 2) 1 Scool of civil engineering, Souteast University, Nanjing, Cina 2 Key Laboratory of Concrete and Prestressed Concrete Structure of Ministry of Education, Souteast University, Nanjing, Cina, ABSTRACT Long-term monitoring data of large-scale civil structures include load-induced and temperature-induced structure responses. Traditional damage detection metods use vibration based structure responses by eliminating temperature effects from te measured data. In tis article, a structural damage identification metod using temperature-induced responses is proposed and applied to a long-span suspension bridge. A structure transfer function is constructed by taking temperature variation and temperature-induced strain as input-output data; tus, it as te potential to accurately reveal inerent structural caracteristics, unlike traditional metods, wic mainly use structural vibration responses from ambient testing. In te proposed metod, te temperature-induced strain is first separated from measured strain responses by using ensemble empirical mode decomposition tecnology; te Euclidean distance matrix is ten defined by using temperature variations and temperature-induced strains for structural damage detection. Numerical simulation and long-term monitoring data of te Jiangyin suspension bridge under normal operating conditions ave been used to verify te proposed metod s effectiveness and robustness. Damage identification of te Jiangyin suspension bridge before and after a sip collision as also been studied wit te proposed metod by using te data from te structural ealt monitoring system. Te researc results sow tat te proposed metod accomplises successful condition and damage assessment. Keywords: Temperature strain; Damage detection; Long-term monitoring; Suspension bridge. 1. INTRODUCTION Civil structures inevitably suffer material deterioration, environmental corrosion, traffic 1 P.D. student, Scool of civil engineering, Souteast University, Nanjing, P.R.Cina 2 *Corresponding autor, Professor, Key Laboratory of Concrete and Prestressed Concrete Structure of Ministry of Education, Souteast University, Nanjing, P.R.Cina, jian@seu.edu.cn

2 loads, and even natural disasters suc as eartquakes and typoons. How to guarantee teir safety and structural ealt as become an important researc area in civil engineering (Aktan et al. 2; Catbas et al. 28). Damage detection is an important tool for structural ealt monitoring and condition assessment. Advanced signal processing tecnologies, including wavelet, ensemble empirical mode decomposition (EEMD), neural networks, and support vector macines, ave been used to develop various kinds of damage detection metods. Vibration-based damage detection metods ave been well developed; tese use te canges in structure vibration caracteristics before and after damage. A variety of damage indices ave been proposed, including mode sape curvature, strain mode sape, flexibility matrix, and modal strain energy; tese mainly use structural acceleration and strain dynamic responses (Brownjon et al. 211; Lei 212; Zang et al. 215). Te limitation of tese vibration-based metods is tat structure vibration caracteristics are generally unavoidably affected by environmental factors, particularly temperature load, during long-term structural monitoring. Accordingly, a number of metods ave been developed to eliminate te temperature effect from structural response, including singular value decomposition, principal component analysis, auto-associative neural network, and support vector macine. However, even if te temperature influence as been completely eliminated, te damage identification results are still not satisfactory because tey only use output data and cannot extract te structural transfer function. Te number of studies of temperature-based structural ealt monitoring as been gradually increasing. Ni et al. (212) studied te correlation between measured temperatures and termal movements of expansion joints for structural cumulative displacement prediction. Kim and Lamam (21) investigated te relationsip between termal load and structure responses, including girder axial force, girder bending moment, pile lateral force, pile bending moment, and pile ead/abutment displacement. Laory et al. (213) used termal variations as load cases to evaluate structural candidate models. Yarnold and Moon (215) proposed a transfer function to calculate structural displacements and restrained member forces from recorded temperature-induced strains. A number of interesting metods are being developed for temperature-based structural ealt monitoring, and it as te potential to update a structural finite element model by using measured temperature load and temperature induced reactions. In tis article, a damage detection metod is proposed using measured temperatures and temperature-induced strains tat were separated from measured strains using te EEMD tecnology, and it is applied to a long-span suspension bridge by using long-term monitoring data including sip collision data. Tis paper is organized as follows: First, te Jiangyin suspension bridge and its structural ealt monitoring (SHM) system are briefly described. Second, te metodology of using measured temperatures and temperatureinduced strains is presented for structural damage detection; tis includes teory derivation for te proposed metod, separating temperature-induced strain by using te EEMD tecnology, a Euclidean distance calculation using measured temperatures and temperature-induced strains, and te damage index defined by te difference between te Euclidean distance matrices at different times. In te tird and fourt sections, numerical

3 simulation and long-term monitoring data of te Jiangyin suspension bridge under normal operating conditions are used to verify te respective effectiveness and robustness of te proposed metod. Damage identification of te Jiangyin Bridge before and after a sip collision is also studied wit te proposed metod by using te monitoring data from te SHM system. 2. THE JIANGYIN BRIDGE AND ITS SHM SYSTEM Te Jiangyin Yagntze River Bridge is a suspension bridge wit a main span of 1,385 meters over te Yangtze River in Jiangsu, Cina (Fig. 1). Te main girder is a welded streamlined flat steel box girder, wit a eigt of 3 m and a widt of 36.9 m, and te navigation clearance is 5 m. It as two 19-m tall reinforced concrete towers. Te suspension cables are ancored directly in rock using gravity ancorages. In recent years, te average traffic flow over te bridge as increased to 7, daily units muc iger tan te 15, daily units of te first year of its operation (1999). Fig. 1 Jiangyin Bridge SHM system A SHM system was installed on te Jiangyin bridge and was upgraded in 25. One undred and sixteen fiber Bragg grating sensors were used for strain(fbgs) and temperature(fbgt) measurement of te main span, including 72 fiber optic sensors on nine equidistant cross-sections of te main span for longitudinal strain measurement, eigt on te mid cross-section for transverse strain measurement, and 36 on nine equidistant cross-sections for temperature measurement as sown in Fig. 1. Tirty-five uni-axial accelerometers (AS) represented by solid circles in Fig. 1 were used in te SHM system, in wic 15 were mounted in positions 1/8, 1/4, 3/8, 1/2, and 3/4 of te main span, eigt on main cables, and 12 on angers. Oter types of sensors suc as GPS and displacement transducers were also used, but tese are not studied in tis article. 3. THE PROPOSED DAMAGE DETECTION METHODOLOGY Termal analysis using temperature as te input as been well studied, and temperature distribution on te bridge is easy to measure. Tus, one can use te measured

4 temperature and temperature-induced response as input and output data for structural identification based on well-developed termal analysis teory. Te framework of te proposed metodology is sown in Fig. 2. Te procedure is described as follows: Fig. 2 Overview of te proposed metodology Step 1: Temperature and structural dynamic strain responses in ambient vibration tests are measured by using te SHM system. Typical temperature and strain time istories of te Jiangyin Bridge are sown in Fig. 2. Tey ave a similar trend, and te temperatureinduced strains are greater tan te traffic-induced strains; Step 2: Temperature-induced strains are separated from measured strains using te EEMD tecnology. Eac location of te temperature-induced strain was used for normalization processing wit temperature. Structural intrinsic caracteristics can be represented by te measured temperature and te temperature-induced strain, tus tey ave te potential for use in structural damage detection; Step 3: Te Euclidean distance of eac sensor is calculated to form te distance matrix by using te monitoring data. Two Euclidean distance matrices at different times are calculated to detect te structure canges between tose times. Step 4: Te damage index matrix, defined as te difference between te Euclidean distance matrices at different times, is calculated. Tis difference matrix will clearly sow weter tere is a cange in eac element of te structure; tus structural damages can be easily located by observing te canges of te damage index matrix as sown in Fig. 2.

5 4. THEORETICAL BASIS OF THE PROPOSED METHOD Te measured temperature is decomposed into uniform temperature,, and gradient temperature, y. Te temperature-induced strain, ε T, for a simply supported beam structure under a uniform temperature cange and te nonlinear temperature gradient is expressed as follows: ε T = ε U + ε N M y + ε y (1) were, ε U = α and ε N y = [ α y b(y)dy]/a are axial strains caused by uniform temperature and nonlinear temperature gradient, respectively. α is te termal expansion coefficient. b(y) is te cross-sectional widt, wic canges wit y. is te cross-section eigt. A is te cross-sectional area. ε M y = [ α y b(y)ydy]y o /I is te temperature induced bending strain at dept caused by te nonlinear temperature gradient. I is te moment of inertia of te cross-section. y o is te distance to te neutral axis. ε U is related to te uniform temperature and te expansion coefficients, ε N y, ε M y, by including te stiffness and M cross-sectional information of structure. Te temperature induced bending strain ε y is separated from Eq. (1). ε y M = ε T ε U ε y N = αt yb(y)ydy I y o (2) It sould be noted tat te temperature-induced bending strain, ε M y, is time varying due to te temperature gradient, y, is time varying. Tus it is not reasonable to make damage M assessment by comparing ε y at different time. As described in Eq. (2), ε M y = [ α y b(y)ydy]y o /I, and α y b(y)ydy in fact is te temperature induced moment. Tus, te bending strain can be normalized to be te output/input format as follows: ε y M = εm y αt y b(y)ydy It is seen tat ε y M is time in-varying and it as direct relation wit structural stiffness. Tus it is able to be used for structural damage detection. Te Euclidean distance of te normalized bending strains (ε n M ) at eac cross-section is expressed as: = y o I d = (ε M ε M ),, = (4) (3) d = d (5) were, d is Euclidean distance between ε M wit ε M. ε M (ε M ) is normalized bending strain of te ( ) cross-section. Wen te strains measured at all sections are used, te Euclidean distance matrix, EDM, at te time t is assembled as

6 d d d d EDM = d d [ d n d n d d n d d n d d n d n d nn ] n n (6) were, d is te Euclidean distance between sensor i and sensor j. n is te section number. Similarly, te Euclidean distance matrix, EDM 1, at anoter time t can also be calculated by using te measured strain at tat time. A damage index is defined as te difference of te Euclidean distance matrices at different times as sown in Eq. (7): EDMD = EDM EDM 1 (7) Te pysical meaning of tis damage index is tat damage occurs wen te Euclidean distance between te normalized temperature induced bending strains at eac section at different times canges. 5. NUMERICAL EXAMPLE OF THE JIANGYIN SUSPENSION BRIDGE Termal analysis of te Jiangyin suspension bridge was performed in ANSYS 14.5 to validate te proposed damage detection metod. Te FE model is sown in Fig. 3. Te girder adopted te elastic sell element 63. Te vertical bending moment of inertia and lateral bending moment of inertia are.844m 4 and 3.3 8m 4 respectively. Te mass density is kg/m 4 wic include dead load and secondary dead load. Te nonlinear element of link 1 is used in te main cable and angers. Te atmosperic temperature at any our t of te day can be expressed by te function: (t) =.5( + n ) +.5( n ) sin [(t ) π/ 2] (8) were, and n are te igest and lowest temperatures in a day respectively. t is te time of day. Te main girder was divided into 18 areas, wic were given different temperature loadings. Te ortotropic bridge deck consists of steel deck plates, aspalt concrete cover, and deck trougs. Te element lengt of te main girder is 1 meter and 18 target elements were abstracted as sown in Fig. 3. Te entity rectangle and ollow rectangular section represent te damage area and intact elements respectively. Te FE model exported te temperature and strain for target elements. Fig. 3 Finite element model and case

7 Difference Value Elements Two damage cases were simulated in te FE analysis. Te case was te multiple damage case in wic te stiffness of te 4 t, 1 t, and 13 t elements were reduced by 1%, as sown in Fig. 3. First, te temperature-induced strain and temperature gradient were extracted. Second, te temperature-induced strain data were normalized by using Eq. (3). Tird, te Euclidean distance matrix was calculated and compared before and after te damage. Wen calculating te Euclidean distance matrices, 1% wite noise was added into te simulated data as observed noise, in wic 1% represents te standard deviation of te standardized bending strain. For te multiple damage case, te calculated damage index for te proposed metod is plotted in Fig. 4. Te damage locations can be clearly identified in bot single and multiple damage cases Elements Elements Elements (a) difference istogram (b) difference matrix figure Fig. 4 Multiple damage identification results (1% stiffness loss and 1% noise) 6. VERIICATION USING MONITORING DATA OF THE JIANGYIN BRIDGE On te nigt of June 2, 25, a piling sip collided wit te second cross-section of te girder of te Jiangyin bridge. Te sensors including accelerometers and fiber optic strain sensors in te SHM system of te bridge monitored girder vibrations during te sip collision. Te collision-induced vibration response lasted about 3 minutes. Te strain amplitude was 53 με at te second cross-section, and was witin 3 με at oter cross-sections. Te strain was very sensitive to te deformation of te bridge. Tus, te real-time strain can be used as an early warning indicator for an emergency event.

8 Difference value Cross section Cross section (a) index difference istogram (b) index difference matrix Fig. 5 Damage detection results before and after te sip collision After te incident, engineering staff were most concerned wit te severity of bridge damage and weter te bridge could continue to be used safely. Te autors ave identified te damage to te bridge by using te monitoring data from 6/1/25 and 6/3/25. If girder damage ad occurred, ten te Euclidean distance matrix would ave canged as a result of te incident. First, te tird and fift sensor strain data were abstracted at eac section by using te EEMD tecnology and ten te bending strain was calculated. Second, te cange of temperature gradient was calculated at eac section and te bending strain was normalized by using Eq. (3). Tird, te Euclidean distance matrices were calculated and compared. Te damage identification results are sown in Fig. 5. Figure 5(a) sows tat tere was no sudden cange of row or column, and no deep color canges witin te istogram. Te maximum Euclidean distance is.4 in Fig. 5(a). It is seen tat tere is no brigt color in Fig. 5(b) wic indicate tere is no significant structural stiffness reduction on te girder due to te sip collision. Visual inspection of te bridge as also been performed, and tere are no obvious damages found. 7. CONCLUSIONS Structural ealt monitoring using temperature-induced responses as received increasing attention from researcers. In tis article, a structural damage identification metod was proposed by using measured temperature and temperature-induced strain data as input-output data, unlike traditional vibration based metods using output-only data. Wit te application of te proposed approac to te Jiangyin suspension bridge, te following conclusions and discussions ave been made: Te input-output relationsip between temperature load and temperature induced response is derived, and te normalized bending strain defined as te bending strain divided by te input term ( α y b(y)ydy, a function of measured temperature) is derived as an indicator for structural damage detection. Te novelty of te propose metod is tat it uses bot te input (te measured temperature) and te output (temperature-induced strain) for

9 structural damage assessment. Numerical studies of a long span suspension bridge successfully verified te effectiveness of te proposed to detect stiffness reduction in a single or multiple locations. Te proposed metod as been performed to evaluate structural safety conditions of te Jiangyin suspension bridge before and after a sip collision, wic illustrates tat no significant stiffness reduction induced by te sip collision. Long term monitoring data as also been studied to evaluate structural safety conditions under normal operation conditions. REFERENCES Aktan, A., Catbas, F., Grimmelsman, K., and Tsikos, C. (2). Issues in Infrastructure Healt Monitoring for Management. J. Eng. Mec., Brownjon, J.W., Stefano, A., Xu, Y.L., Wenzel, H., and Aktan, A.E. (211). Vibrationbased monitoring of civil infrastructure: callenges and successes. Journal of Civil Structural Healt Monitoring, 1, Catbas, F.N., Susoy, M., and Frangopol, D.M. (28). Structural ealt monitoring and reliability estimation: long span truss bridge application wit environmental monitoring data. Eng Struc., 3(9), Kim, W., and Laman, J. A. (21). Integral abutment bridge response under termal loading. Eng Struc., 32(6), Lei, Y., Jiang, Y.Q., and Xu, Z.Q. (212). Structural damage detection wit limited input and output measurement signals. Mec. Syst. Signal Process., 28, Laory, I., Westgate, R.J., Brownjon, J.M.W., and Smit, I.F.C. (213). Temperature variations as load cases for structural identification. Te 6t international conference on structural ealt monitoring of intelligent infrastructure, Hong Kong. Ni, Y. Q., Xia, H. W., Wong, K. Y., and Ko, J. M. (212). In-Service Condition Assessment of Bridge Deck Using Long-Term Monitoring Data of Strain Response. J. Bridge Eng., ASCE, 17(6), Rezaei, D., and Taeri, F. (29). Experimental validation of a novel structural damage detection metod based on empirical mode decomposition. Smart Materials and Structures, 18(4). Wu, Z.H., and Huang, N.E. (29). Ensemble empirical mode decomposition: a noise assisted data analysis metod. Advance in Adaptive Data Analysis, 1(1), Xu, Y. L., and Cen, J. (24). Structural damage detection using empirical modes decomposition: Experimental investigation. J. Eng. Mec., ASCE, 13(11), Yarnold, M.T., and Moon, F.L. (215). Temperature-based structural ealt monitoring baseline for long-span bridges. Eng Struc., 86, Zang, J., Ceng, Y.Y., Xia, Q., and Wu, Z.S. (215). Cange Localization of a Steel Stringer Bridge troug Long-Gauge Strain Measurements. ASCE Journal of Bridge Engineering /(ASCE)BE