Research On The Visualization Method Of Special-shaped Metal Parts Based On Statistical Electromagnetic Imaging

With the development of social economy and the progress of industrialization,metal parts have been widely used in industrial field.In order to meet the requirements of technology and production,special-shaped metal parts with irregular shape are developed.Various types of defects are inevitably appeared on the surface or inside the special-shaped metal parts,e.g.automobile parts,during manufacture and use.Hence the defect detection methods should be improved to avoid the potential safety hazard.As a visual electromagnetic detection method,Electromagnetic Tomography(EMT)technology has the advantages of no radiation,non-contact,and high sensitivity,which provides an effective solution for the detection of special-shaped metal parts.Visualization can directly display the detection results in the form of images.In order to improve the quality of reconstructed images,sensor design and image reconstruction algorithm are studied in this paper.The special-shaped metal parts of automobile is studied in this paper.An“O”-shape sensor array is designed according to the structure of the metal part,the sensor simulation model is optimized,the relationship between the parameters of sensor coil and the EMT image quality is analyzed and discussed.The EMT system is constructed for experiment according to the optimized sensor parameters.Image reconstruction of EMT is a typical nonlinear,ill-posed and ill-conditioned inverse problem.The traditional regularization algorithms for EMT reconstruction are only to find a single estimate,hence the amount of model information provided is limited.A large number of reasonable model parameter estimates can be obtained by statistical methods.This paper proposes an image reconstruction algorithm based on Bayesian theorem for EMT.The posterior distribution model of the dielectric constant of EMT is obtained based on Bayesian theory and priori information of standard normal distribution.In order to extract local features of defect image,the conductivity distribution is divided into a series of block structures according to the sparsity of defect distribution.With the aid of sparse Bayesian learning(SBL)framework,statistical information,including the prior probability of the conductivity sparse distribution and the noise information in the measurement data,is taken into account.Hence the full statistical description of the conductivity distribution can be obtained.Finally,Defects of special-shaped metal parts are imaged by using sparse Bayesian algorithm with optimized sensor simulation model in this paper.In order to verify the effectiveness of the algorithm,the new method is compared with the Tikhonov regularization algorithm,the Total Variation(TV)regularization algorithm and the EM probability statistics algorithm.Both simulation and experimental results show that the relative errors of images reconstructed based on the Bayesian algorithm with statistical information can be reduced by 20% compared with traditional methods.As a result,the quality and accuracy of defects images are effectively improved.