Study On Key Technology Of Health Monitoring Based On Strain Modal For Bridges

Bridge is an indispensable part of the traffic engineering which is close with thenational economy, the people's life and property and the stability of the society. However,the collapsed and damaged accidents of bridges frequently emerged at home and abroad inrecent years lead to more people's questions about the safety of bridges. If the healthcondition of bridges was known and some measures have been adopted to fix those damagedbridges before collapse, a lot of life and property will be saved. Therefore, the healthmonitoring technologies of bridges that make bridges becomes more intelligent and the safeof bridges can be efficiently controlled become the necessary research direction of bridgeworkers either now or in the future.This paper relies on National High Technology Research and Development Program("863Program") of China (Project No.2009AA11Z104) and health monitoring technicalresearch project of An Longquan overpass in National Road102. Based on strain modaltheory, these technologies that sensor optimal placement, parameter identification anddamage detection are studied from the structural dynamic characteristics. The geneticalgorithm(GA), differential evolution(DE) and fuzzy clustering(FC) method are adopted inthe course of the study. There are some innovation in every key technology research. In thestudy of sensor optimal placement, a new fitness function is designed based on modal energycriterion(BHM) and effective independence criterion(EI) or modal assurance criterion(MAC).Moreover, the two-step method for sensor placement is proposed learning from stabilitydiagram(SD) theory and the sensor placement commonly used for girder bridges is presented.In the course of parameter identification research, the SD based on FC method is proposed todistinguish the spurious modes in which the frequency as the abscissa and any column dataof MAC as the ordinate. Furthermore, the DE algorithm is used to determine the clusteringcenters of SD that greatly improves the success rate and accuracy of SD with FC. In theresearch of damage detection technology, the DE method is also put forward to study themultiple damage location of bridges. The two-step method is proposed through the control ofdamage range, that firstly to determine the damage location and then the damage extent. Theresearch contents almost cover a series of process of health monitoring from sensorplacement to damage detection. The improved new methods and technologies further guarantee the safety of bridges and intelligentize the no life bridges by a big step forward.The specific researches in this paper are as follows:1. The theoretical formula of strain response are deduced according to the relationbetween strain and displacement modal. As the displacement modal, it is found that the strainmodal can be expressed by modal superposition theory. The orthogonality among strainmodal mode shapes is proved and the strain modal coordinate is same with the displacementmodal coordinate of the same order. The two advantages that strain modal is better than thedisplacement modal is verified. One is avoiding the transition error of additive process fromdisplacement to strain, the other is increasing the local sensibility of structures. According tothe result of error analysis, the transition error is small for bridge project. However, theaccumulative error will affect the accuracy if the transition error adds on the measure orother inherent error of the displacement modal in practical engineering. The study for thebiggest advantage of strain modal has found out that strain modal is more sensitive to localchange and damage than displacement, and thus more conducive to bridge healthmonitoring.2. The Single Parents Genetic Algorithm(SPGA) is a kind of simple and effectivesensor optimal algorithm because that there is no the crossover operation and then keeps thenumber of sensors unchangeable in the evolutionary process. Five fitness functions aredesigned based on different criterions. Two-step method is proposed to determinate thenumber and the location of sensors refering to the stability diagram theory. The analysis forthe number of sensors indicates that the number and location of sensors depend on the orderof mode shape selected. The number range of sensors can be determined by the turning pointof curve N-f. In the curve, N is the number of sensors and f is the fitness function f2and thelog value of the fitness functions f1and f4. The analysis results of the location of sensors areas follows: The locations placed by f2based on MAC are disperse and difficult to form theideal stable poles. Whereas the locations placed by f3based on BHM concentrate onside-span. Therefore, the locations of sensors can be determined by the comprehensiveconsideration for f1, f4and f5. In addition, the locations commonly used for simple beam andcontinuous beam are given: the best locations for simple beam are the midspan section, thenare the sections near1/4and3/4span. The optimal locations for the continuous beam arefirstly near the midspan, sencondly are the middle supported sections of midspan.3. The stochastic subspace identification(SSI) that only needs ambient excitation isapplied to the parameter identification of bridge structure. The comparative results between covariance-driven stochastic subspace identification(Cov-driven SSI) and data-drivenstochastic subspace identification(Data-driven SSI) demonstrate that the Cov-driven SSI getsmore higher identified prescion and much less runtime than Data-driven SSI. The traditionalstabilization diagram(TSD) is improved by fuzzy C-Means clustering(FCM) and the inputspace fuzzy kernel clustering method (KFCM). The stabilization diagram based on FC isproposed in which the frequency as the abscissa and any column data of modal assurancecriterion (MAC) as the ordinate without the consideration of damping. The compared circlemethod is presented to distinguish the spurious and physical modes so that the identifiedprocess of parameters becomes more intelligent. The DE method is used to initialize theclustering centers in order to improve the success rate of SD based on FCM and KFCM. Dueto the combination of DE and KFCM-â…¡, the identified results of modal parameter becomemore accurate.4. The identification of multiple damage locations is carried on based on DE methodafter the theory analysis of damage identification. Multiple damage location assurancecriterions(MDLAC) based on frequency or frequency and mode shape are designed. Thetwo-step method is presented to identify the mulple damage locations after determinating theindividual, damage range, damage scale and evolutionary generation. The fist step is to findout the suspected damage locations. The second step is to determine accurately the damagelocations and extent for suspected damage locations. The research results demonstrate thatMDLAC based on frequency can't determine the accurate locations for symmetricalstructure. Whereas the symmetrical elements including the elements damaged can beidentified. Therefore, the mode shape should be introduced into MDLAC based on frequencyin order to determine the locations accurately. The identified results with MDLAC based onboth frequency and mode shape in the first step is superior to MDLAC only based onfrequency. The two-step method still can determine the damage locations and extentaccurately after the introduction of noise. It is proved that the identification method ofmulple damage locations based on DE has the stronger anti-interference ability.