Research On State Evaluation Based On The Similarity Of Thermography Principal Curves In Manifold Space

Eddy current pulsed thermography(ECPT)is a new electromagnetic nondestructive testing technology,which integrates the advantages of eddy current testing and infrared thermography.It is widely applied in detection and evaluation of defects in the industry and is one of the development trends of non-destructive testing technology in future due to the advantages of non-contact,high detection efficiency,wide detection range and no interference to the original temperature field and so on.However,existing ECPT techniques are insensitive to early damage,especially micro-pattern defects that have not yet formed cracks.Moreover,the existing ECPT methods mainly analyze and process thermal images of a single frame or a specific moment,which are lack of analysis of the relationship between mathematical separation models and physical models and the relationship between features and material macro characteristics.From the perspective of multi physics fusion characteristics of ECPT,this paper proposes an ECPT processing method based on laplacian eigenmap in manifold space and the similarity of principal curves to detect and evaluate material state according to electrical,magnetic,thermal and other physical effects of material.The proposed method establishes the relationship between material macroscopic properties parameters and the defects.The simulation and experiment are conducted to validate the feasibility of the proposed method.Firstly,this paper analyzes the detection system,basic principles of ECPT and influential factors,and introduces the processes involved in ECPT including electromagnetic induction,joule heating,heat conduction and thermal radiation.According to the mechanism and characteristics of ECPT,the spatial-transient-phase thermography is selected for analysis.In view of the nonlinearity and high coupling of the space-transient-phase thermography,laplacian eigenmap algorithm is adopted to obtain the low dimension embedded manifold of high dimensional thermography.The main curve of low dimensional manifold is extracted,and the detection and recognition of material defects are carried out by the comparison of the similarity of the main curve.Secondly,Simulation for eddy current pulsed thermography is performed in COMSOL Multiphysics.The distribution of the magnetic field,eddy and temperature of the specimen is obtained.On this basis,the thermal imaging information under the different material property parameters is obtained by changing the macroscopic characteristic parameters of the material.The simulation thermal image is processed by the similarity evaluation method of principal curves in manifold space,and the relationship between the characteristics and the material macroscopic characteristic parameters is established.Finally,this paper validates the proposed defect detection method based on the similarity assessment of principal curves in low dimensional space.The method was used to detect the fatigue micro-damage of gears,and the damage evolution process under different fatigue stress was analyzed.