Detection Of Bubble Defects In Plywood Based On Vibration Acoustics

Based on the acoustic characteristics of vibration,the vibration or sound generated by transient excitation on plywood is recorded.Extraction and analysis of the relevant characteristic parameters of the acoustic response signal of plywood under vibration excitation.Genetic algorithm is used to optimize the neural network model to detect and locate the bubbling defects of plywood.The main contents and research results of this paper include the following aspects:(1)The acoustic response signal acquisition system of transient excitation vibration of plywood was constructed.In order to improve the repeatability and accuracy of signal acquisition,the parameters of the acquisition system were compared and optimized.The basic conditions and parameters of the response signal acquisition system are determined as four-point support of the plywood sample,fixed acceleration sensor with magnetic seat,hammer head made of rubber material and 10 k Hz signal sampling frequency.(2)Pre-processing methods such as starting point correction,maximum amplitude normalization and Butterworth low-pass filter denoising for plywood vibro-acoustic signals are used to improve the analysis contrast between samples and reduce the error of signal characteristic parameters.The vibro-acoustic signal of plywood is analyzed in the time and frequency domains,the characteristic parameters in the time and frequency domains are extracted,and the influence of different plywood layers,the number of bubbles and the adhesive layer,the direction of vibration propagation,and the excitation and detection points on each characteristic parameter and correlation are studied.Twelve characteristic parameters,such as peak-to-peak value,root mean square value,kurtosis,shape indicator,crest indicator,impulse indicator,clearance indicator,signal attenuation coefficient,center of gravity frequency,root mean square frequency,frequency standard deviation and fundamental vibration frequency,are optimized to construct the basic plywood bubbling recognition feature vector.The bubbling defect location has the greatest influence on the kurtosis value of the three-layer and five-layer plywood,and the average variation amplitude is 11.09% and 5.56%,respectively.The root-mean-square frequency of seven-layer plywood has the greatest influence,and the average variation range is 4.14%.The number of bubbling defects has the greatest influence on the margin factor of the third layer and the seventh layer,and the average variation range is 19.07% and25.49%,respectively.The kurtosis value of five-layer plywood has the greatest influence,with an average variation of 15.03%.(3)The acoustic response signal of plywood vibration is analyzed by wavelet packet,and the energy spectrum of wavelet packet components with different wavelet basis functions and decomposition layers are compared and analyzed.Using the db4 wavelet basis function to decompose the signal with four-layer of wavelet packets,the energy ratio of the first six components has a more obvious effect on the detection of plywood bubble defects.(4)Using BP neural network as the basic model of plywood bubble defect recognition,genetic algorithm is introduced to optimize the model,and a GA-BP neural network model for plywood bubble defect recognition is construction.Taking the constructed plywood bubble recognition feature vector as the input parameter of the model,and the number of bubbles and the number of layers as the output parameters,the recognition accuracy of the basic model of plywood bubble defect recognition BP neural network and the GA-BP neural network model were compared.The results show that the recognition accuracy of the BP neural network model is 81.82%,and that of the GA-BP neural network model is 90.91%,which is significantly improved.