Thesis (Ph.D.)--Chulalongkorn University, 2007

The identification of multiple vehicle dynamic axle loads on multi-span continuous bridge is proposed. The objective is to develop a practical technique to determine dynamic axle loads of multiple vehicles based on measured bridge responses. Following the concept of inverse problem of returning bridge responses into time-varying point loads, the solution can be determined using least squares regularization optimization via singular value decomposition (SVD) method. The updated static component (USC) technique is adopted to improve the accuracy and eliminate the difficulty of an optimal regularization selection. The computer simulation and experimental studies are conducted to investigate the effectiveness of the proposed method. A small scaled model of a three-span, continuous bridge and two scaled 2-axle vehicles are designed and fabricated. Study of identification accuracy affected by related parameters including measurement noise, system modeling, vehicle mass, moving speed, axle spacing, number of vehicle axle, bridge surface roughness as well as various moving schemes of multiple vehicle travel are considered. The actual dynamic axle loads of the vehicle models during their travel are directly measured and used in accuracy evaluation. From the obtained results, it is observed that the employed technique effectively improves the accuracy and robustness of the identified loads, particularly around the internal bridge supports in which the problem of zero identified loads are eliminated. Vehicle mass, moving speed, axle spacing and bridge surface roughness significantly affect to identification accuracy. A heavy, widely spaced axle vehicle traveling with slow speed on smooth surface bridge yields better accuracy of identified axle loads than a light, closely spaced axle vehicle moving with fast speed on rough surface bridge. In addition, it is found that the method is robust and accurately identifies dynamic axle loads for all moving schemes of vehicles. No conflict of the identified axle loads during axle overlapping or passing the bridge support is observed because the axle loads are independently identified and controlled by static influence lines from the USC algorithm. The comparison between the measured and reconstructed bending moments indicates that the approach is correct. The accuracy of identified dynamic axle loads for all cases of study is within relative percentage error of 13%. Moreover, axle load identification with incomplete measurement information is investigated and recommendations toward the real application of the proposed identification system are also given.