Research On Audio Detection Method Of Carbide Saw Blanking

The hidden defect of carbide saw blade is a major hidden danger of product quality and safety.How to detect implicit hidden defects online has been concerned and puzzled by manufacturing enterprises.There are many methods of nondestructive testing.In this paper,audio lossless detection technology is used to explore the internal recessive defects of the carbide saw blade.In this paper,the conventional nondestructive testing technology is preliminarily studied.The new audio detection technology and its development status are mainly analyzed.By analyzing the characteristics of the audio signal of carbide saw blade,the audio detection technology is applied to carry out the research on its signal.Then lists the common defects of hard alloy saw blade and the analysis of the shortcomings of conventional NDT methods,analyzed the principle and theory of audio detection technology,audio signal of hard alloy saw analysis pretreatment and time-frequency analysis experiment,and put forward a kind of realize the endpoint detection of hard alloy saw audio signal using the average amplitude method.Study on the extraction of audio features hard alloy saw blade defect,HHT(Hilbert-Huang Transform),LPCC(Linear Prediction Cepstrum Coefficient)and MFCC(Mel Frequency Cepstrum Coefficient),their analysis of the hard alloy saw audio signal extraction ability,and proposes a method of extending HHT mean extension,improve the parameters of HHT the characteristics of hard alloy saw blade audio extraction model.Finally,in order to realize the automatic identification of hard alloy saw blade without defect,this paper established the HHT parameter as input of BP neural network model,realizes the automatic classification of carbide saw audio signal by using BP neural network method,this paper after several tests and elastic gradient descent method achieves convergence requirements.