An ensemble neural network for optimising a CNC milling process

Computer numerical control (CNC) milling is a common method for the efficient mass production of products. Process efficiency and product quality have a strong dependency on the cutting process conditions. Furthermore, optimising a process for material removal rate (MRR) and surface roughness (SR), which are measures of process efficiency and product quality, respectively, is a complex optimisation task due to their contrasting relationships with process parameters. In this work, CNC end milling is performed on aluminium 6061 to investigate the influence of key process input variables (feed per tooth, cutting speed, and depth of cut) on the machined part's SR. Firstly, a full factorial parametric study is conducted and analysed using Analysis of Variance (ANOVA) before an Ensemble Neural Network (ENN) is trained on the experimental data. To capture the complex nonlinear relationships accurately, each base model of the ENN is a combined genetic algorithm-artificial neural network, whose hyperparameters are optimised using a Bayesian optimisation framework. Once trained, the ENN predictive model is exploited to identify optimal input parameter permutations to achieve a predefined SR value while maximising MRR. Analysing the experimental data demonstrates that the SR performance envelope is nonlinear with respect to the input variables. Furthermore, the ANOVA results indicate that feed per tooth is the dominant input parameter with a contribution of 40.2% and that there are strong interactions between the input parameters investigated. The ENN performance was subsequently validated through a further set of experiments producing a mean absolute percentage error in predicted SR of just 2.56%.