Mathematical Modeling Using ANN Based on k-fold Cross Validation Approach and MOAHA Multi-Objective Optimization Algorithm During Turning of Polyoxymethylene POM-C

Abstract

The paper has a dual purpose: firstly, to examine the influence of various cutting conditions (cutting speed 𝑉 𝑐 , feed 𝑓, depth of cut 𝑎𝑝, tool nose radius 𝑟 ɛ , and cutting edge angle 𝑋 𝑟 ) on the quality of machined parts (𝑅𝑎), tangential force (𝐹 𝑍 ) and cutting power (𝑃 𝑐 ) during the turning process of polyoxymethylene POM-C. Two carbide inserts, SPMR 120304 and SPMR 120308, were used for the three-dimensional cutting operations. Secondly, the goal is to identify optimal cutting conditions that maximize material removal rate (𝑀𝑅𝑅) while minimizing three output parameters (𝑅𝑎, 𝐹 𝑍 , and 𝑃 𝑐 ). The study employed analysis of variance (ANOVA) to assess the significance of the input parameters on the desired outcomes and utilized an artificial neural network (ANN) to create mathematical models. The K-fold Cross-Validation approach was deemed suitable due to its efficiency in requiring fewer experiments. To optimize the cutting conditions, a new metaheuristic optimization algorithm called Multi-Objective Artificial Hummingbird Algorithm (MOAHA) was selected. ANOVA analysis reveals that factors 𝑓 and 𝑟 ɛ contribute 58.05% and 32.25%, respectively, to the response 𝑅𝑎. Classical parameters (𝑉 𝑐 ,𝑓, and 𝑎𝑝) also impact mechanical cutting actions (𝐹 𝑍 and 𝑃 𝑐 ).The MOAHA algorithm, coupled with four ANN models, optimized the five cutting conditions, resulting in optimal values 𝑉 𝑐 = 250 𝑚/𝑚𝑖𝑛, 𝑓 = 0.08 𝑚𝑚/𝑟𝑒𝑣, 𝑎𝑝 = 1.3 𝑚𝑚, 𝑟 ɛ = 0.8 𝑚𝑚, and 𝑋 𝑟 = 75°. Under these conditions, responses are: 𝑅𝑎 = 0.6 µ𝑚, 𝐹 𝑍 = 21.51 𝑁,𝑃 𝑐 = 60.24 𝑊, and 𝑀𝑅𝑅 = 26.38 𝑐𝑚 3 /𝑚𝑖𝑛. The ANN-MOAHA coupling provides an excellent, simple, and fast computer tool for multi-objective optimization.