A machining error tracing method based on MEA-BP neural network for quality improvement of gear hubs

Abstract

ABSTRACT The machining quality of workpieces is greatly influenced by the performance of an equipment. Furthermore, it is difficult to establish an error tracing model with high tracing accuracy using a mathematical method. In this study, the machining quality of gear hubs for an automobile synchronizer produced on an intelligent manufacturing line was evaluated. The main sources of machining errors were analyzed, and the machining error tracing model for the gear hub was established through a back propagation (BP) neural network. To improve the performance of the error tracing model, the weights and thresholds of the BP neural network were optimized using the mind evolutionary algorithm (MEA). The MEA-BP error tracing model was trained and tested using online measurement results and historical data of the production line. The results showed that the average tracing accuracy of the MEA-BP method was 97.4%, which was 12.1% higher than that of the BP method. The average running time of the MEA-BP was far less than that of a genetic algorithm (GA) improved BP method. These comparisons prove that the proposed MEA-BP error tracing method is both feasible and effective. The proposed method can improve the machining quality and error tracing in intelligent manufacturing applications.