Research On Moving Force Identification Of Bridge Based On Elastic Network Regularization

It is very important for bridge safety assessment,structural fatigue life estimation and management and control of over-limit and over-load to obtain the moving force accurately.In practical engineering,it is difficult to measure the moving force directly,so the indirect identification has been widely concerned.The identification of moving force is the second kind of inverse problem in structural dynamics,which is usually ill-conditioned.Regularization is a common method for solving ill-conditioned problems.In this paper,a new method of moving force identification based on elastic network regularization is proposed,which is verified by numerical and experimental results.Here’s how it works:(1)The research status of moving force identification methods is summarized.The principle of the method,the well-posedness of the recognition equation,the recognition accuracy,the noise robustness and sparsity of the recognition results are compared.Aiming at the regularization recognition method which has the highest versatility,the paper points out the shortcomings of the existing methods and determines the focus and main research contents of this paper.(2)A regularized moving force identification method based on elastic network is proposed.The proposed method is a combination of Tikhonov regularization method and L1 norm regularization method,which can effectively reduce the oscillation of the result caused by the measurement noise,and thus obtain a stable approximate solution whose identification accuracy meets the requirements,while retaining the ability of obtaining sparse solution,thus,the moving load characteristics can be preserved on the premise that the recognition accuracy is basically unchanged.The basic flow chart of moving force identification is described,and the theoretical feasibility of the proposed method is demonstrated.(3)A numerical example is used to verify the effectiveness and feasibility of the proposed method.The discrete trigonometric function is selected to form the load dictionary to realize the sparse representation of load.The influence of moving speed,noise level.combination of measuring points,load time history and weighting coefficient on the proposed method is studied by using simple beam model under single moving load and double moving load respectively.The results show that the proposed elastic network regularization method can identify moving loads and is robust to noise.(4)A comparative study is made between the methods presented in this paper and those of least squares.Tikhonov regularization method,11 norm regularization method.The finite element model of the test beam under double moving load is established,and the pre-test is carried out at the same time.The influences of noise level,measuring point location and bridge support form on different identification methods are compared.Results the validity of the proposed method is verified,and it has the advantages of Tikhonov regularization method and L1 norm regularization method.(5)The experimental case expansion method is used to verify.A single span and single box girder model was built in the laboratory,and the vehicle-bridge coupling dynamic experiment was carried out.Based on the experimental data,the moving force model adopts single moving force and double moving force,and the test beam adopts elastic support beam.The experimental results show that the proposed method is effective and the identification accuracy is high under the elastic support beam working condition.