Uncertainty quantification in machining deformation based on Bayesian network

Abstract

Uncertainty quantification in the analysis of machining systems is of great importance for continuously improving product quality, reliability, and efficiency of manufacturing processes. This paper presents a novel method for quantifying the influence of uncertain factors on machining deformation. Initially, uncertainties are evaluated using the method of moment estimation and least squares method for autoregressive models, deemed prior information. Then, a Bayesian network for machining deformation is established. Finally, all prior information is imported into the Bayesian model and an algorithm is used to compute the posterior probability. The influence of residual stress on machining deformation was taken as an example, and a detailed analysis was carried out. Our findings highlight the uncertainty of machining-induced residual stress (MRS), which was found to vary from 0.12 to 0.36, and the uncertainty of initial residual stress (IRS), which varied from 0.18 to 0.53. Furthermore, the presence of machining-induced residual stress increased the probability of machining deformation from 1.0% to 6.4%; while initial residual stress can increase the probability of machining deformation by up to 17.8%. For other factors such as material properties, workpiece geometry and stiffness of the machining system, the total combined influence of uncertainties on machining deformation was 9.1028E-04. The results highlight the importance of quantifying the effect of uncertainties on machining deformation.