Research On Crack Testing System Based On Pulsed Eddy Current

The Research of Crack Testing System Based on Pulsed Eddy Current proceeds with principle of eddy current testing and according to above,the technology of achieving pulsed eddy current testing signals is mainly to be discussed.After having finished removing the noise form the signals with wavelet transform and achieving crack information,the signals eigenvalues are extracted with principle component analysis method,crack classification algorithm is chose to dispose the signals further according to eigenvalues.Artifical neural networks and support vector machine algorithm are mainly researched to classify with the achieving eigenvalues and to identify surface and sub-surface crack of metal materials,so the method can accurately grasp reliability of metal materials and provide effective basis for using and updating the metal materials further.The application of the crack detecting algorithms applying to Crack Testing System Based on Pulsed Eddy Current is mainly discussed,including the classification of detecting signals and comparative analysis of the two algorithms in the paper.BP neural networks adopt gradient descent method to learn and nonlinear differential function to train the sample data,it has the abilities of well global approximation and generalization.Modified Generic Algorithm with real instead of binary data is adopted to help optimize artifical neural networks in global domain for finding suitable weights and thresholds as quickly as possible and completing the training of BP neural networkds,so MGA-BP algorithm are formed.Support Vector Machine algorithm is a effective classified method based on statistical learning theory,it has the great advantage of classification for small sample space,it can transform original sample space high-dimension sample space making use of optimal and generalized optimal hyperplane to classify the sample into two categories or more,it has better generalization capability.Models are built with Matlab for BP neural networks and SVM algorithm are used to test signals of pulsed eddy current and analyse the results.It makes sure that BP neural networks is used to implement the classification of the testing signals can improve the recognition rate of crack in Crack Testing System Based on Pulsed Eddy Current.Crack Testing System Based on Pulsed Eddy Current is designed with DE2 development platform,FPGA component and SOPC designing technology.The design including implementing software and hardware of crack detecting algorithm, choosing component of input and output system,configing and testing of the system is mainly completed.To test the detecing data of long bolt with the system as test platform,it chieves the better result.