Design Of Piston Throat Near-surface Micro Flaw Detection Probe And Study Of Defect Recognition

The piston is one of the most important parts in the engine.The throat of the piston is the first key bearing area in the working process of the piston,which directly contacts with the fuel in the working process.Its near surface quality has a direct impact on the performance and life of the engine.With the full implementation of China VI emission standard,in order to ensure the full combustion of fuel in the engine cylinder to meet the new emission standard,the piston needs to bear the burst pressure of more than 20MPa.When the piston throat has a fine defect,the explosive force will lead to the defect cracking,resulting in serious safety risks,so the research on the piston throat fine defect detection is of great significance.In order to solve the difficulty of detecting the near surface defect of piston throat,On the basis of previous research,the finite element simulation model of eddy current testing system based on the near surface defect of piston throat was established by using COMSOL simulation software,to study the influence of various parameters of the eddy current testing probe on the detection sensitivity.According to the simulation results,the probe parameters are determined as follows:Cylindrical excitation coil height 4.5mm,Coil diameter 1.5mm,the number of turns of coil 1338.Other detection parameters such as excitation frequency 10000Hz,lift-off height 0.46mm.Because the eddy current detection signal of the piston throat near the surface of the micro defect contains many noise signals,and the signal-to-noise ratio(SNR)is low,in addition,for the signal noise reduction with unknown real value,the traditional evaluation methods for noise reduction are one-sided.An improved wavelet threshold criterion is proposed on the basis of previous studies,the negative number of SNR,root mean square error and smoothness are weighted by the coefficient of variation method,the quality of signal denoising is measured by a new comprehensive evaluation method.The parameters of wavelet denoising are determined based on the comprehensive evaluation index,finally,Sym8 is chosen as the wavelet function to decompose the eddy current detection signal in 4 layers,and the proposed improved threshold criterion is used for soft threshold denoising.In order to distinguish the types of micro-defect signals near the piston throat,firstly,the wavelet packet energy spectrum characteristics of each signal after denoising were extracted based on wavelet packet decomposition,and the extracted feature values were used to train the improved active learning classifier based on SVM.Select Pistons of the same type and different production batches to experiment,among them,cutting damage defect type samples(M),casting damage defect type samples(C)and non-defect piston samples(N)were marked with 100 groups each,80 groups(L)and 500 unlabeled samples(U)of various defect types were randomly selected to form the training set,the rest of the sample is the test set(T).In the two experiments,the stop criterion was reached after 8 and 7 iterations respectively,the overall accuracy of sample identification of the test set was 93.33%and 95%,and the accuracy of defect identification was 96.67%and 100%,respectively.