Real-time thermal error compensation on machine tools using improved BP neural network

Abstract

A BP neural network was improved to solve the problem of thermal error compensation for the in-feed axes thermal deformation of CNC machine tools. First, the rough sets theory was used to analysis the correlation between all measuring data and thermal error, and sorted out key characteristic data for the thermal error compensation of machine tool. And then, an artificial neural network with a dynamic feedback network was put forward to set up the thermal error compensation model and integrated in the open architecture control system of th e actual machine. The new thermal error real-time compensation method could produce higher error compensation accuracy, faster convergence speed of online learning and better real-time performance of the error compensation in CNC machine tool Simulation results showed the feasibility and validity of this method.