The Application of Hybrid BPNN and GA in the Dry Machining End-Milling Process of Aluminum Alloy

The environmentally friendly practice of dry machining protects worker health, saves cutting tools, and has fewer adverse environmental effects. A hybrid approach has been employed during the dry-end milling process of aluminum alloys Al 6061 for predicting and optimizing multi-performance characteristics, being minimum tool flank wear (VB), low Surface Roughness (SR), and maximum Material Removal Rate (MRR). The varied end milling process variables are the operation type (up milling and down milling), spindle speed (n), feed speed (Vf), and radial depth of cut (ar). The experimental design employed is a Taguchi L18 orthogonal array. The quality characteristics of VB and SR are smaller-the-better, while the quality characteristic of MRR is larger-the-better. The initial step is to model the end milling process and to predict the responses using a Back Propagation Neural Network (BPNN). The next step is to acquire the suitable levels of the machining variables that would give optimum performance using a Genetic Algorithm (GA) method. The third step is calculating the percentage of the end milling process variables' variance to the total variance of the responses (percent contribution or PC) and analyzing the response charts. The confirmation experiment indicates the combined BPNN and GA could estimate and enhance the multi-performance features greatly and accurately. The type of operation, i.e., down milling, has the highest PC value for VB and SR, while Vf has the highest PC value for MRR. The results and conclusions of the investigation's work can be applied immediately to knowledge-based and expert systems for dry milling applications in industrial practices and for further study into sustainable production and design.