Deep learning-driven optimization and predictive modeling of LASER beam machining for XG3 steel

Area of Science:

• Manufacturing Engineering
• Materials Science
• Additive Manufacturing

Background:

• LASER Beam Machining (LBM) is a precise, non-contact thermal process for advanced materials.
• It's crucial for aerospace and defense due to minimal mechanical stress on difficult-to-cut alloys.
• XG3 steel is a high-performance alloy requiring optimized machining for critical applications.

Purpose of the Study:

• To experimentally investigate and optimize LASER Beam Machining (LBM) for XG3 steel.
• To evaluate the influence of cutting speed, gas pressure, focus point, and depth of cut on machining outcomes.
• To develop and compare predictive models for LBM process performance.

Main Methods:

• Conducted experiments using a Taguchi L27 orthogonal array for XG3 steel.
• Varied parameters: cutting speed (8-12 m/min), gas pressure (0.5-0.9 Bar), focus point (2-6 mm), depth of cut (3-9 mm).
• Employed Analysis of Variance (ANOVA), Multi-Objective Genetic Algorithm (MOGA), Response Surface Methodology (RSM), and Back-Propagation Artificial Neural Network (BPANN).

Main Results:

• Cutting speed was the most significant factor, influencing surface roughness (>82%), machining time (>74%), surface hardness (>81%), and burr thickness (>84%).
• MOGA identified Pareto fronts for optimal trade-offs, achieving surface roughness of 1.10-1.16 μm and machining times of 2.44-2.52 s for circular profiles.
• The BPANN model demonstrated superior accuracy (R > 0.999) compared to RSM, with low MAPE for surface roughness (1.48%) and hardness (0.72%).

Conclusions:

• LASER Beam Machining parameters, particularly cutting speed, critically affect XG3 steel's machining quality.
• Optimized LBM parameters and MOGA provide effective solutions for balancing multiple performance objectives.
• BPANN is a highly accurate predictive tool for LBM processes, outperforming RSM.