Application of machine learning regression algorithms to optimization of CMM measurement strategies

Abstract In this paper, optimization of coordinate measuring machine (CMM) measurement strategies using machine learning regression algorithms is investigated. The study analyzes the impact of key strategy parameters (mainly the measurement speed) on measurement accuracy and repeatability by using the Zeiss Contura equipped with an active probe head. Experimental data are used to develop regression models for optimizing the CMM strategy towards limitation of measurement deviations. Currently, parameter selection relies on the manufacturer’s recommendations and operator input. The novelty of this research lies in applying machine learning techniques to automate this process. Three different regression algorithms—linear regression (LR), support vector regression (SVR) and decision tree regression (DTR)—are evaluated using a dataset consisting of 60 measurement points. In addition, polynomial feature transformations are applied to input data in order to capture nonlinear relationships. Models are assessed using 10-fold cross-validation and performance metrics such as the mean absolute error and the root mean squared error. The results indicate that polynomial transformations improve model performance. LR achieves the lowest errors and the best balance between accuracy and model complexity. SVR shows strong generalization, while DTR exhibits overfitting. Learning curves and residual plots confirm that regularization improves decision tree performance. The obtained outcomes demonstrate that machine learning algorithms can effectively optimize CMM measurement strategies, reducing operator dependency and enhancing efficiency. Future work should expand the dataset and explore additional algorithms to refine prediction accuracy and parameter selection.