Separation Of Bridge Dynamic Deflection Using Sparse Regularization Under Multi-time Scale

With the rapid growth of the number of newly-built bridges in China,the healthy operation of bridges in service is gradually receiving attention and attention.Health monitoring of the structure of the bridge in service can not only timely evaluate the safety state and performance of the structure,but also predict the safety risks of the bridge structure,so as to take necessary measures in advance to prevent the bridge collapse and other emergencies.Among them,the dynamic deflection of the bridge is an important parameter reflecting the health of the bridge structure.Since the dynamic deflection of a bridge is the deflection value caused by various effects and factors,in order to accurately evaluate the working performance and health status of an active bridge,the deflection contribution caused by the main factors in the dynamic deflection of the bridge need to be extracted separately,such as vehicle load,temperature difference effect,temperature effect caused by the deflection and the loss of prestress.Although some scholars have tried to separate the dynamic deflection of bridges,most of them remain in the theoretical level and the separation effect is not good.In order to solve this problem,this paper proposes a sparse representation based separation method for bridge dynamic deflection under multi-scale system analysis.The main research contents are as follows:Firstly,the actual engineering problems of bridges are transformed into mathematical problems,and a mathematical model of bridge dynamic deflection separation is established.By introducing the sparse regularization model and taking into full account the signal characteristics of annual temperature difference deflection,daily temperature difference deflection and long-term deflection,the joint dictionary is constructed to make the linear combination of the elements in the dictionary describe the bridge deflection signal as much as possible.In order to obtain the sparse representation of dynamic deflection,the L1 norm regularization term is selected.Since the L1 norm regularization is a convex optimization problem,and its solution does not generally have an explicit format,this paper proposes to use a fast compression threshold value iterative algorithm to solve the above problem to obtain the component deflection.Secondly,in order to verify the feasibility and effectiveness of the theory presented in this paper,a finite element model was established for numerical simulation.Combined with the numerical simulation of bridge dynamic deflection under multiple time scales,the discrete resolution,atomic participation number,long-term deflection,annual(day)temperature difference effect accuracy,and long-term deflection model of polynomial functions that may affect the separation structure of bridge dynamic deflection during sparse representation And analysis of signal length and other parameters,and dynamic deflection decomposition using singular value decomposition and eigenvalue analysis under the same conditions,comparing the results obtained under different separation methods,and combining the dynamic deflection signal separation performance evaluation indicators to find the most Excellent sparse separation method.Thirdly,the dynamic deflection data of two adjacent spans of a bridge were collected on site,and a MIDAS calculation model was established according to the actual parameters of the bridge.The collected dynamic deflection data were preprocessed by the data cleaning method combined with sparse representation theory,and the abnormal data such as default value and mutation value existed in the measured data were corrected.The method proposed in this paper is used to separate the measured data after pretreatment,and the feasibility and applicability of this method are further verified,which lays a theoretical foundation for the practical application of this method.The results show that the method of separating bridge dynamic deflection based on sparse representation can effectively separate the daily temperature difference deflection,annual temperature difference deflection and long-term bridge deflection of the signal,and the separation result has higher accuracy.Different parameters have different effects on the separation performance of bridge dynamic deflection signals.Whether the main characteristics of the atom constructed by the dictionary can well simulate the corresponding signal law is the key to affect its sensitivity to parameters.