Prediction of tool wear based on GA-BP neural network

Abstract

The anisotropy and nonuniformity of wood-plastic composites (WPCs) affect the milling tool, which rapidly wears during high-speed milling of WPCs. Thus, the evolution mechanism of tool failure becomes complicated, and the prediction of tool wear cannot be precisely described mathematically. A neural network based on tool wear test was proposed to predict the tool wear condition during high-speed milling of WPCs. The traditional backpropagation (BP) neural network easily falls into the local optimal solution. A genetic algorithm (GA-BP) neural network prediction model was established by using the GA to optimise its initial weight and threshold. The BP model and the GA-BP model were evaluated in terms of mean square error and training times, and the generalisation verification was applied to the prediction model. After analysing and comparing the results of the two models, the GA-BP neural network model has better training speed and accuracy under the test conditions. The relative error between the predicted value and the actual value is controlled within 5%.