Intelligent Acoustic Detection Of Anvil Crack For Cubic Apparatus

The synthetic diamond is primarily produced in China,which commonly applies the static synthesis technique by using large-volume cubic apparatus.Due to the high pressure,high temperature and alternative stresses in the production,cracks often occur in the carbide anvil,thereby resulting in significant economic losses or even casualties.Conventional methods are unsuitable for crack detection of the carbide anvil.This paper proposed a novel intelligent acoustic detection method of carbide anvils using the support vector machine and deep learning technique.The specific research contents and results are as follows:(1)Aiming at the problem of crack representation under complex background noise,an adaptive feature extraction algorithm for anvil crack based on signal energy and PCA technique is proposed.According to the energy threshold of acoustic signal,a sliding window interception method is used to extract the independent sound pulse from the origin acoustic signal.By comparing and analyzing the statistical characteristics of cracked and normal sound pulse,the feature matrix composed of zero-crossing rate,sound pressure level and linear prediction cepstrum coefficient is established.Then,PCA technique is introduced to eliminate redundant information in original feature.The simulation results show that proposed method can effectively characterize state of the anvil.(2)A crack detection method for anvil based on SVM-kNN is proposed.The initial SVM model is established by using grid optimization and cross validation techniques.The sigmoid function is introduced to calculate the posterior probability of the output for SVM,and then the reliability interval of SVM prediction results is obtained.For the suspected fault samples within the interval,a kNN classifier is designed to make a secondary classification.The experimental results show that the SVM-kNN model has high r-ecognition accuracy.(3)Aiming at the problem of insufficient generalization ability of artificial feature extraction,and complex mapping relationship between anvil state and signal cannot be represented by shallow network structure,an intelligent detection method of anvil crack based on SAE-PSO is proposed by introducing deep learning technology.Firstly,the sliding window and FFT technology are used to build the data set for deep learning of anvil crack.According to the signal reconstruction error and stochastic gradient descent algorithm,a three-layer SAE initial diagnosis model is established.An improved PSO algorithm is proposed and used to optimize Dropout and weight decay coefficients of the model.The experimental results show that compared with SVM,PCA-SVM and SAE methods,SAE-PSO algorithm not only has the highest recognition accuracy,but also effectively improves the generalization ability of the model.