Prediction of Surface Roughness of 304 Stainless Steel and Multi-Objective Optimization of Cutting Parameters Based on GA-GBRT

Abstract

Establishing and controlling the prediction model of a machined surface quality is known as the basis for sustainable manufacturing. An ensemble learning algorithm—the gradient boosting regression tree—is incorporated into the surface roughness modeling. In order to address the problem of a high time cost and tendency to fall into a local optimum solution when the grid search and conjugate gradient method is adopted to obtain the super-parameters of the ensemble learning algorithm, a genetic algorithm is employed to search for the optimal super-parameters in the training process, and a genetic-gradient boosting regression tree (GA-GBRT) algorithm is developed. A fitting goodness of fit is taken as the fitness function value of the genetic algorithm and combined with k-fold cross-validation, as such, the initial model parameters of the gradient boosting regression tree are optimized. Compared to the optimized artificial neural network (ANN) and support vector regression (SVR) and combined with the cutting experiment of 304 stainless steel with a micro-groove tool, a genetic algorithm multi-objective optimization model with the highest cutting efficiency and a supreme surface quality was constructed by applying the GA-GBRT model. The response relationship reveals the non-linear interaction that occurs between the cutting parameters and the surface roughness of 304 stainless steel that is machined by the micro-groove tool. As indicated by the results obtained from the multi-objective optimization, the cutting efficiency can be enhanced by increasing the cutting speed and depth within a small range of surface quality variations. The GA-GBRT model is validated to be reliable in making a prediction of the surface roughness and optimizing the cutting parameters with turning and milling data.

Multi-Objective Optimization of Cutting Parameters in Turning AISI 304 Austenitic Stainless Steel

“…For example, Kumar [4] adopted surface roughness and material removal rate (MRR) as objectives to optimize the cutting parameters in turning C360 copper alloy. Zhou et al [5] obtained the Pareto optimal solution with the maximum MRR and the minimum surface roughness in turning AISI 304 based on the genetic algorithm gradient boosting regression tree (GA-GBRT) model they established. Their experimental results demonstrated that MRR can be improved by increasing cutting depth and cutting speed in a small range of surface roughness variations.…”

A Review of Indirect Tool Condition Monitoring Systems and Decision-Making Methods in Turning: Critical Analysis and Trends

“…Besides, according to the results of the study, they determined that the input that affects the SR the most are the feed rate and then the axial vibration. Zhou et al [ 154 ] created an algorithm based on the genetic-gradient boosting regression tree for the optimization of cutting parameters and SR estimation by turning AISI 304 stainless steel. They compared this model created with an optimized artificial neural network and support vector regression method.…”

Comparison of Surface Roughness When Turning and Milling

“…Zhou et al [15] investigate the turning process, where the cutting speed, the feed and the depth of cut were varied. The surface roughness prediction model was built to the measured data and use genetic-gradient boosting regression tree method.…”