Study On Multimodal Ultrasonic Guided Wave Fusion For Estimating The Thermal Stress Of Continuously Welded Rail

Continuous welded rail is widely used in high-speed railway and urban rail transit.Since the rail cannot expand or contract freely in the longitudinal direction,there will be a huge thermal stress inside the rail when the rail temperature changes.In severe cases,it may lead to accidents such as rail buckling and broken rails,threatening the operation safety directly.Therefore,it is of great significance to monitor the internal longitudinal thermal stress of continuous welded rail and give warning early before the stress exceeds the limit of the safe operation of high-speed railway.In this thesis,the problem of longitudinal thermal stress detection for continuous welded rail is discussed and the basic theory and key technical issues in rail stress detecting based on the guided wave are studied systematically.An estimation method based on the multimodal guided wave fusion is proposed to realize the accurate estimation of the longitudinal thermal stress of the rail under the conditions of unknown elastic modulus.This result is of great theoretical and practical significance for the development of related technologies.Based on a semi-analytical finite element rail model,the multimodal characteristics of guided wave in rail are analyzed under different stresses and elastic modulus.The deficiencies and limitations of the existing stress detection methods based on the guided wave are summarized.Based on these,a series of key issues to be solved urgently are proposed.The optimal methods are studied for selecting,exciting and receiving the optimal modes.And then the group velocity and phase velocity of target modes are calculated by using Hilbert transform and two-dimensional Fourier transform respectively,and a double weighted correction method is proposed.Based on this,the relationship between the stresses and the velocities of guided wave modes in different frequencies are established.The accuracy of stress detection is then analyzed based on the group velocity or the phase velocity.The results show that,with the increase of the guided wave frequency,the calculation error of the group velocity increases gradually,while the error of the phase velocity is less affected.Therefore,in practical applications,it is necessary to choose an appropriate type of velocity as the detection parameter to estimate the rail stress according to the guided wave frequency.A multimodal fusion stress estimation algorithm based on Support Vector Machine is proposed to solve the problems of low accuracy and poor anti-interference ability using single mode.The method uses 2D Fourier transform to obtain the characteristic signal of the guided wave in the rail,and the two-dimensional energy spectrum of each single mode is excited.The relationship function of the two-dimensional Fourier transform spectrum between the single mode and the longitudinal stress is established using a support vector regression model.The numerical simulations show that the estimation standard deviation of the proposed algorithm using 200 Hz guided wave modes is less than 0.05 MPa in the ideal condition.A stress estimation algorithm based on multimodal ultrasonic guided wave fusion is proposed to solve the problem that existing stress detection methods based on waveguide are greatly affected by the elastic modulus of the rail.The algorithm based on different sensitivities of modal wavenumbers to the longitudinal thermal stress and the elastic modulus,and the selection of modal minimum error criteria and the neural network estimation model are proposed.The input feature vectors of the model are multimodal wave numbers and rail temperature,and the output vector of the model is the longitudinal thermal stress value.The model can overcome the influence of the changing elastic modulus on the longitudinal thermal stress detection.Simulation experiments show that the standard deviation of the estimated rail thermal stress of the model using 35 kHz guided wave modes is less than 1 MPa with 5%white noise added.Based on theoretical analysis,various mathematical modeling methods and a large number of numerical simulations,modal excitation/reception methods and sensitive parameter acquisition for the arbitrary guided wave modes in the rail are studied.A multimodal combination optimization and parameter fusion algorithm is proposed to accurately estimate the longitudinal thermal stress of the rail under the conditions of unknown elastic modulus.It can adapt to the actual conditions in the long-term service of continuous welded rail,and provide a solution to the problem of longitudinal thermal stress detection for continuous welded rail.