Research On Evaluation Method Of Eddy Current Thermal Imaging Based On Hidden Markov Model

Eddy current pulsed thermography is a kind of thermal thermography nondestructive testing technology with electromagnetic induction eddies as the heat source.It has the advantages of non-contact,non-interaction,high detection efficiency,and intuitive detection information,etc.It is widely used in scientific research and social production.The current methods for defect detection and analysis of eddy current pulsed thermography usually select one or a few frames of thermal image at a specific time,and less analysis is conducted from space,time and structural state,combining physical process and mathematical model.In this paper,based on the multi-physical field coupling process of eddy current pulsed thermography,the multi-frame thermal image sequence data is firstly preprocessed.Then the Hidden Markov Model with statistical learning and probabilistic reasoning ability is introduced into data processing of eddy current pulsed thermography.Finally,the effectiveness of the proposed method is verified by finite element simulation and experiments,which provides a new research idea for damage detection and evaluation of metal components.In this paper,the damage detection mechanism of eddy current pulsed thermography,various influencing factors and the experimental system of eddy current pulse thermal imaging are introduced.The characteristics of thermal image signal and the thermal state of damaged specimen are analyzed,and the theory of Hidden Markov Model is expounded.Based on the characteristics of thermal image signal,the thermal image sequence is converted into a multidimensional observation matrix with specific physical meaning.The algorithm of Laplacian Eigenmaps is used to reduce the data dimensions of the multidimensional observation matrix and extract the low-dimensional features of the thermal image.By means of the method of interval sampling of thermal image,the sample of thermal image is expanded while the frequency of the sequence of thermal image is kept certainly high.Finally,the reduced low-dimensional observation matrix with multiple samples is input into the Hidden Markov Model to realize the evaluation of different damage states of materials.COMSOL finite element simulation software is used to establish the electrical,magnetic and thermal multi-physical field simulation model,and the multi-physical field simulation App is developed.In the simulation process,the multi-physical process of eddy current pulsed thermography is analyzed.The evolution of the conductivity,thermal conductivity and magnetic permeability of materials is evaluated by using the Hidden Markov Model.Finally,the proposed method is used to evaluate the fatigue damage of the gear and the stress state of the tensile specimen.The experimental results are consistent with the fatigue evolution process of the gear and the deformation law of the material under force.Simulation and experimental results show that the proposed method is effective.