Research On Stress Identification Method Of Steel Strands Based On Support Vector Regression

Steel strands are the key stressed components of various types of long-span girder bridges and rigid frame bridges,which directly determines the bearing capacity and durability of the structure.However,affected by environmental erosion,material creep,load overrun and other factors,the performance degradation of steel strands inevitably occurs during the long service period of the bridge,which reduces the long-term performance and safety of the structure.Carrying out the research on the stress monitoring/detection technology of steel strands in service bridges,and proposing accurate and effective stress monitoring/detection technology for bridges is conducive to establishing a more perfect bridge state assessment system,which is of great significance for ensuring the safe operation and scientific management of the increasingly large "aging" bridges in China.In this paper,the research on the stress level detection method of steel strands is carried out.The entry point of this study is that the ultrasonic guided wave propagation characteristics of steel strands are affected by the stress state.The finite element simulation and experimental methods are used to analyze the guided wave propagation process of steel strands under different tension conditions.A stress feature extraction algorithm based on wavelet packet transform and singular value decomposition is proposed to extract the stress feature parameter-singular value vector of guided wave signal.The Euclidean distance and support vector regression method is introduced to process the singular value vector of guided wave to realize the quantitative identification of the stress level of steel strands.The main research results are as follows:(1)A stress feature extraction algorithm based on wavelet packet transform and singular value decomposition is proposed to process guided wave signals.The effectiveness of the algorithm is verified by using finite element data and experimental data.The singular value vector extracted from the guide wave signal can reflect the stress change,and the feature dimension of the singular-value vector is less,which is convenient for the further quantitative identification of the stress level of steel strands.(2)A finite element model of guided wave propagation in steel strands with different stress levels is established.The influence of the application mode of tension load,amplitude curve and application duration on the stress distribution after tension of steel strands,as well as the influence of friction coefficient and excitation signal on the mode and waveform of guided wave are discussed.Using the most suitable finite element model,the propagation process of guided wave in steel strands with different stress levels is simulated,and the guided wave signal is extracted.(3)A method for identifying stress levels of steel strands based on singular value vector distance is proposed.The feasibility of the method is verified by finite element simulation and experimental research.In addition to the large change in the singular value vector distance from the unstressed state to the 0.1Rm stress state,the singular value vector distance has a good linear relationship with the stress level,and the finite element results are consistent with the experimental results.(4)An optimization method for identifying stress levels of steel strand based on support vector regression is proposed.Taking the singular value vector as input and the stress level of the strand as the output,the support vector regression model suitable for stress recognition can be obtained by training the model,which can realize the stress level identification of the steel strands.A particle swarm optimization algorithm is used to optimize the model parameters,and the applicability of the method is confirmed by comparing the stress identification results of the neural network model.