Multiobjective Optimization of Roll Materials via Data Analytics-Mechanism Integration

Abstract Laser cladding provides a promising route for low-carbon remanufacturing of cold-rolling rolls. However, the strong coupling among process parameters makes precise control of coating geometric characteristics challenging. Using 45 steel as the substrate, we constructed a data set of process parameters and coating geometric characteristics. An evolutionary learning model based on an artificial neural network (ANN) optimized by an improved differential evolution (IDE) algorithm achieved an average coefficient of determination (R2) of 0.9230 on the test set for predicting three coating geometric characteristics. Under engineering constraints, nondominated sorting genetic algorithm II (NSGA-II) was applied for inverse multiobjective optimization, yielding the optimal parameters: a laser power of 1200 W, a scanning speed of 233.2 mm/min, and a powder feed rate of 2.89 r/min. Experimental validation showed prediction errors below 10%. This study provides a validated closed-loop framework for intelligent process optimization in cold-rolling roll remanufacturing based on the 45 steel/T15 M material system.