Abstract
LASER Beam Machining (LBM) has emerged as a highly precise and non-contact thermal machining process, widely adopted for cutting complex geometries in advanced engineering materials. Its ability to machine difficult-to-cut alloys with minimal mechanical stress makes it particularly suitable for aerospace and defense components. This paper presents an experimental investigation and multi-objective optimization of LASER Beam Machining (LBM) for XG3 steel, a high-performance alloy used in aerospace and defense applications. The study evaluates the impact of four process parameters i.e. cutting speed (8, 10, 12 m/min), gas pressure (0.5, 0.7, 0.9 Bar), focus point (2, 4, 6 mm), and depth of cut (3, 6, 9 mm) on four output responses: surface roughness, machining time, surface hardness, and burr thickness. Experiments were conducted using a Taguchi L(27) orthogonal array on three distinct hole geometries: circular, triangular, and square. Analysis of Variance (ANOVA) revealed that cutting speed was the most dominant factor, contributing over 82% to the variation in surface roughness, 74% for machining time, 81% for surface hardness, and 84% for burr thickness. The interaction between cutting speed and depth of cut was also found to be statistically significant. For single-objective optimization, the ideal parameters to minimize surface roughness were a cutting speed of 12 m/min, gas pressure of 0.5 bar, focus point of 2 mm, and depth of cut of 3 mm. Multi-objective optimization using a Genetic Algorithm (MOGA) generated Pareto fronts to identify balanced trade-off solutions; for a circular profile, this resulted in surface roughness values of 1.10-1.16 μm and machining times of 2.44-2.52 s. Furthermore, two predictive models, Response Surface Methodology (RSM) and a Back-Propagation Artificial Neural Network (BPANN), were developed. Comparative analysis showed the BPANN model was significantly more accurate, with regression coefficients (R) exceeding 0.999 and Mean Absolute Percentage Error (MAPE) values of 1.48% for surface roughness and 0.72% for surface hardness, confirming its superior predictive capability.