Research On Intelligent Extracting And Analyzing The Features Of Defect On Metal Surface

With the rapid development of national technology and industry,the requirement of magnetic flux leakage detection technology on metal surface defects detection have become increasingly,the test results need to be able to give precise extent of damage.Magnetic flux leakage detection will generate a huge amount of data on metal surface defects for sure,and the core of magnetic flux leakage detection technology is to calculate defect size accurately based on the detected magnetic flux leakage data.Therefore,in order to improve the accuracy and efficiency of metal corrosion detection,it is significant to find out the appropriate method of extracting and analyzing the defect data.This thesis mainly studies and analyzes the metal magnetic flux leakage data detected from metal pipelines,and mainly completes four aspects of work:complete the detection of metal surface defects,achieve the identification of defect signal features,accomplish the examination of features extraction algorithm and realize the analysis of defect signal features.At first,a defect adaptive threhold detection algorithm is studied.Above all,the detected magnetic flux leakage signals are smoothed to remove interference.Then defects of the smoothed magnetic flux leakage signals are detected by the proposed adaptive threhold detection algorithm,and the defect detection results are analyzed.Second,the identification of defect signal features is completed.The feature quantity related to the defect size is found by observing and comparing magnetic flux leakage signals of different size defects.The basic feature quantity is extended so that it can more fully reflect the information of defect signal,and achieve the full identification of the defect features.Third,an intelligent defect features extraction algorithm is studied.In order to realize the accurate extraction of defect features,an intelligent extraction algorithm for defect features is designed based on Matlab software platform and wavelet transform method.The proposed features extraction algorithm is used to extract the features of different defects,and through the analysis of the extracted results,the imperfect part of the algorithm is found and improved,so that the intelligent features extraction algorithm of the defect features is completed.Fourth,a defect features analysis algorithm based on PCA is designed.To start with,the feature quantities related to the length,width and depth of the defect are found through control variate method.After the classification of defect features,a sample database for defect feature analysis is established.The original features are simulated by using the PCA-based defect features analysis algorithm,and the optimization of defect features is achieved.Fifth,a defect features analysis algorithm based on KPCA is designed.To overcome the shortcomings of the PCA analysis algorithm,the original defect features are analyzed by using the KPCA-based defect features analysis algorithm.The dimension of the defect features is reduced,the new features with strong correlation with the defects are found,and the analysis results of two methods are compared.Finally,based on the summary of the thesis,the future research direction is forecasted.