Analysis method for evaluating uncertainty in the machining process of thin-walled parts

The machining process of thin-walled components is full of many uncertainties, resulting in problems such as high batch inconsistency and low pass rate. In this paper, the Bayesian network uncertainty inference model of machining process is constructed. The influence mechanism of input variables on machining distortion uncertainty is clarified. Initially, the uncertain variables are collected and the inference model of machining distortion is constructed based on root-branch-leaf Bayesian network structure. The weight of the influence of each input variable on machining distortion uncertainty and maximum influence path are obtained. Next, an inverse Bayesian network is established, with the uncertainty inference results used as prior information to carry out inverse inference of machining distortion uncertainty. The influence possibility of the related factors of the main influence variables on the uncertainty of machining distortion is obtained. Aviation aluminum alloy T-shaped part was taken as an example, and the influence mechanism of initial residual stress, surface residual stress, cutting force and their related factors on the machining distortion uncertainty was investigated. The results indicated that the influential weights of initial residual stress, surface residual stress and cutting force on the machining distortion uncertainty was 0.33 (maximum), 0.23 and 0.04 respectively. The factors related to initial residual stress, surface residual stress, and cutting force had influence weights of 0.099 (maximum), 0.009, and 0.001 on the machining distortion uncertainty. The probabilities of the effects were 0.920, 0.075 and 0.005, respectively. Finally, the paper compares the proposed model with MC-GBRT and BiLSTM-UP, showing average improvements of 46.7% and 63.1% in stability and 35.7% and 20.6% in computational speed, respectively. This paper proposes a Bayesian network-based machining distortion uncertainty inference model that effectively reveals the mapping mechanism between initial residual stress, surface residual stress, cutting force, and the machining distortion uncertainty in thin-walled components.