Speaker
Description
Abstract
Additive manufacturing (AM) has emerged as promising manufacturing technology for producing geometrically complex components with reduced material waste and greater design flexibility. However, surface quality of additively manufactured components remains an important concern because surface roughness can affect dimensional accuracy, functional performance, and the structural integrity of fabricated parts. Therefore, accurate prediction of surface roughness is essential for improving process control and achieving the desired quality of AM components. This study presents a supervised machine learning-based framework integrated with explainable artificial intelligence (XAI) techniques for predicting the surface roughness of additively manufactured polylactic acid (PLA) specimens. Nine regression algorithms, namely Support Vector Regression, Random Forest, XGBoost, AdaBoost, CatBoost, Decision Tree, Extra Tree Regressor, Explainable Boosting Machine, and Gradient Boosting Regressor, are investigated and comparatively evaluated. The predictive capabilities of these algorithms are assessed to determine their effectiveness in establishing relationships between the considered process parameters and the resulting surface roughness. In addition to evaluating prediction accuracy, XAI techniques are incorporated to improve model transparency and provide meaningful insights into the influence of individual input parameters on predicted surface roughness. This approach addresses an important limitation of conventional machine learning applications, in which high-performing models are often treated as black-box predictors with limited interpretability. Comparative results demonstrate that XGBoost achieves the best predictive performance among the investigated algorithms, attaining a coefficient of determination (R²) of 0.9634. This result indicates that the XGBoost model can effectively capture the nonlinear relationships between input parameters and surface roughness of AM PLA specimens.
| Topic Area | Design Optimization, Additive Manufacturing Technologies & Applications |
|---|