Reliability analysis of thermal error model based on DBN and Monte Carlo method

Abstract

A physically-based thermal error model of the servo axis was established, aiming at the time-varying nonlinear thermal error of the screw. However, the thermal characteristic parameters of the model may differ from the working state of the machine tool. In order to analyze the influence of the variation of the thermal characteristic parameters on the prediction results of the model, and considering that the performance functions of the physically-based servo axis thermal error model are implicit and have no explicit analytical expression, a new method for calculating model reliability using deep belief network (DBN) and the Monte Carlo method was presented, and DBN was used to substitute the implicit functions. The reliability and residual prediction of the model with single parameter and multi-parameter fluctuations was determined. Finally, the robustness of the model and the accuracy of the proposed reliability calculation method were experimentally verified.