The Research Of Structural Damage Identification Based On Wavelet-Extreme Learning Machine Algorithm

With the progress of technology and industry,complex structure has become the development trend of the world.These complex structures are particularly vulnerable to the effects of various uncertainties in the environment and will inevitably result in various damage accumulation.The undetected damage may change the strength and stiffness of the structure,which leads to the accumulation of massive damage and the sudden failure of the structure eventually.People have to face economic and mental loss because of structural damage.Therefore,it is of great theoretical significance and engineering value to carry out health monitoring and damage identification to prevent the risk of structural failure.Wavelet analysis is an analytical method in time domain and frequency domain,which can effectively recognize the singularity of signal.Therefore,the damage location of the structure can be identified according to the location of the wavelet singular point in the signal image.The extreme learning machine algorithm is a new neural network.It only needs to set the number of hidden nodes,the network weight value and the bias of the hidden layer of the algorithm are randomly determined,which is not adjusted during the whole training process.In the end,the only optimal solution is obtained.In this paper,the advantages of wavelet analysis and extreme learning machine algorithm are combined,and a structural damage identification method based on wavelet-extreme learning machine algorithm is proposed.The main contents of this thesis are as follows:1.Wavelet analysis and extreme learning machine algorithm are used to study structural damage identification.The two-stage identification structure damage pattern is adopted in this paper.The principle of wavelet analysis and extreme learning machine is given in detail.The method of structural damage identification for wavelet-extreme learning machine algorithm is established.This method effectively compensates for the shortcomings of wavelet analysis and greatly improves the computational efficiency of damage identification.2.The singularity of wavelet transfrm is used to identify the location of structural damage.Finite element analysis software is used to establish the structural model of damage and obtain the modal parameters.Wavelet coefficients are obtained by wavelet transform of modal parameters.It can be seen that the position of mutation in the map is the location of structural damage.3.The extreme learning machine algorithm is applied to the recognition of structural damage degree.A nonlinear mapping relationship between the maximum value of wavelet model and the degree of damage can be generated by the extreme learning machine training.This relationship can be used to identify the degree of structural damage.The least square method of extreme learning machine is superior to gradient descent method of traditional neural network in selecting and adjusting parameters.So that its learning speed and network generalization are good.4.In order to verify the feasibility of the wavelet-extreme learning machine algorithm in identifying the damage,the study was conducted on the beam and frame structures with multiple injuries.Numerical simulation results show that the method is effective and accurate for small damage identification.Therefore,the wavelet-extreme learning machine algorithm established in this paper has certain reference value for engineering application.