Prediction of Surface Topography of Laser Interference Textured Steel Using Machine Learning and Surfalize

Understanding how laser process parameters influence surface topography is crucial for precise laser surface texturing. While the complex relationship between laser process and topographical parameters is difficult to model analytically, it lends itself well to machine learning. The requirement for large datasets of topographic parameters has generated a need for software solutions based in Python and equipped with batch functionality. In this work, we demonstrate the application of the self-developed Python library Surfalize to analyze a large dataset of direct laser interference patterning textured surfaces in terms of roughness parameters and train different machine learning models to predict topographical features from process parameters. The results show that both the random forest regressors and gradient boost machines exhibit the best predictive accuracy across a wide range of parameters, reaching R² values above 0.9 for amplitude related features such as the structure depth and arithmetic mean height. On the other hand, k-nearest neighbors and support vector machines perform significantly worse. Moreover, parameters from the functional family are predicted with less accuracy than amplitude or hybrid parameters.