Bending Process Parameters Optimization of the 316L Tube with Small R/D Ratio: A Physical Prior Knowledge Based Bayesian Optimization Approach

Dingyang Zhang,Zili Wang,Shuyou Zhang,Jie Liu,Yiming Zhang,Jianrong Tan

2023-01-01

Abstract:In the metal tube plastic bending process, the processing parameters have a significant impact on the final bending quality, which is usually determined by establishing a finite element model for simulation. Processing defects such as wrinkles of metal tubes become more severe as the bend radius decreases and the diameter of the tube increases. These defects can affect the proper functioning of the metal tube or even render it unusable. Existing methods for the selection of metal tube bending process parameters are often based on experience and lack algorithmic guidance, resulting in poor performance in terms of machining result and limits. We propose an improved Bayesian optimization algorithm for tube bending process parameters based on a prior knowledge guidance of physics experiments to optimize the bend processing parameters. In this method, the theoretical study of the changing trend of pipe bending processing parameters is taken as prior knowledge, and the selection of parameters is guided according to the influence trend of processing parameters on defects. The experiments on 316 stainless steel tube bending show that our method can speed up convergence. Additionally, the final optimization results of our algorithm are significantly better than those of traditional Bayesian optimization algorithms.