Prediction of Geometric Characteristics of Laser Cladding Process by the GWO-BPNN Algorithm

• • Polynomial regression model predicts clad width and height with relative error ≤4.2%, enabling reliable pre-selection of process parameters and reducing trial-and-error costs in industrial cladding operations. • • GWO-BPNN model achieves average R² of 95.28% for dilution rate, forming coefficient, and wetting angle, a 12.4% improvement over conventional BPNN (82.93%), significantly enhancing prediction accuracy for multi-objective quality optimization. • • Full-factorial experiments on 316L stainless steel with 316L powder validate the model's robustness across varying laser power, powder feed rate, and scanning speed, confirming its applicability for dynamic process control. • • Inverse validation demonstrates stable predictive performance within engineering tolerances, providing a quantitative foundation for real-time parameter adjustment and quality assurance in laser cladding production lines.