Damage identification and failure characterization of plywood based on AE-BP Model

Abstract

The objective of this study is to improve the accuracy of damage identification of plywood boards by the approach of utilizing acoustic emission (AE) in conjunction with a backpropagation (BP) neural network model and elucidate the failure characteristics under varying working conditions. Six AE characteristic parameters were collected simultaneously at the time of loading test. The K-means clustering analysis method was used to describe the damage evolution process of plywood. Based on the correspondence between the damage degree and the AE characteristic parameters, the damage identification model was established using the BP neural network. The results demonstrated that AE parameters analysis is capable of effectively drawing the distinctions between three damage stages during the stress damage process. The proportion of shear failure of plywood is higher than tensile failure. K-mean cluster analysis revealed a strong correlation between damage types and AE peak frequency. The backpropagation neural network model is subjected to rigorous testing and training. The results show that the model has excellent performance in damage type identification. Therefore, the joint AE-BP model was found to be a considerably effective method to evaluate damage types for plywood products.