Modeling and optimization of the declineation angle generated by abrasive waterjet multi-pass and forward angling cuttings using machine learning

This paper introduces a modeling and optimization study on declineation angle (DA) produced by abrasive waterjet (AWJ) multi-pass and forward angling cuttings with machine learning algorithms (MLAs) for the first time. In this regards, several rock samples such as marble, travertine, basalt, onyx, and tuffs were pre-dimensioned and subjected to AWJ cutting tests. Then, the DA was measured for each rock type and the recorded data was utilized for modeling and optimization by various MLAs. Additionally, the proposed models’ robustness was evaluated using several statistical indices. Moreover, a random forest regressor was employed to determine the relative importance of predictor variables influencing the DA. It was found that the multi-pass and forward angling cuttings have superiorities for reducing the DA. In addition, the artificial neural network, gradient boosted decision trees, and Gaussian process regression were determined as the first, second, and third best models in modeling the DA respectively. Furthermore, the most significant variables influencing the DA were identified as the number of pass and traverse speed. Finally, it was found that the particle swarm optimization algorithm is a successful tool for establishing the optimum variable design.