Research on optimization method of nanosecond laser paint removal parameters based on machine learning

With the growing demand for high cleanliness and precision in metal surface treatment in modern industry, traditional paint removal methods have proven inefficient, environmentally unsustainable, and prone to substrate damage, failing to meet contemporary industrial standards. In contrast, laser cleaning technology has emerged as a promising alternative due to its superior efficiency, precision, and environmental compatibility. This study proposes a machine learning-based framework for predicting the quality of laser-induced paint ablation. An experimental platform employing an infrared nanosecond laser was developed to collect data on paint removal performance under varying process parameters, forming the basis for the machine learning dataset. By integrating support vector regression, artificial neural networks, and random forest (RF) algorithms, nonlinear mapping models were established to correlate process parameters with cleaning quality metrics, including removal rate, surface roughness, and paint removal thickness. Bayesian optimization was utilized for parameter tuning. Experimental results demonstrate that the single-objective RF model achieves exceptional predictive accuracy, with a test set R2 value of 0.9465 (single-objective model). Building on these findings, an intelligent optimization framework combining RF with Bayesian optimization was developed, enabling the identification of high-performance parameter combinations. The optimized predictions yielded a removal rate of approximately 98.9%, a paint removal thickness of around 100.8 μm, and a surface roughness of about 2.74 μm. This work integrates multialgorithm modeling with intelligent optimization strategies, providing a robust theoretical and technical foundation for the intelligent control of laser cleaning processes.