Ensemble machine learning model for predicting surface roughness during drilling of hybrid sisal-cotton fiber reinforced polyester composites

Abstract

Composites that integrate natural fibers with polymer matrix possess versatile properties and profoundly contribute to sustainability and eco-friendless. Machining of these composite materials is challenging due to material’s anisotropic property, non-homogeneous structure and abrasive nature of fibers. Surface roughness measures the quality characteristics of drilled hole on hybrid sisal-cotton reinforced polyester composite materials. To predict surface roughness, researchers have used empirical and statistical techniques, both of which have issues with generalizability when forecasting unseen data. The objective of this work is to build a super-learner machine-learning model to predict surface roughness of drilled hole on hybrid sisal-cotton reinforced polyester composite using a variety of machine-learning techniques, such as decision trees, random forests, gradient boosting, and extreme gradient boosting. K-fold cross-validation and grid search hyperparameter optimization are used to optimize the models. The proposed super learner model’s prediction performance is contrasted with the other models. With a coefficient of determination (R 2 ) of 99.4% between the experimental and predicted values for surface roughness for the test dataset, the proposed super learner model outperformed the other models in terms of prediction. With the highest coefficient of determination (R 2 ), lowest mean absolute error (4.75%), mean absolute percentage error (3.30%), and root mean square error (5.32%), this model is shown to be the most accurate. Additionally, the Shapley additive explanation technique, which highlights the crucial elements influencing the surface roughness of polyester matrix composites reinforced with sisal and cotton fibers, explain the model’s predictions. In the end, a tool based graphical user interface has been developed for the machine learning prediction model’s practical use. This software predicts surface roughness quickly, accurately, and intelligently.