Explainable Artificial Intelligence (XAI) and Machine Learning Technique for Prediction of Properties in Additive Manufacturing
Abstract
Explainable artificial intelligence method is used to understand the underlying phenomenon of the machine learning (ML) algorithm prediction. In this work, a powerful XAI technique, SHapley Additive exPlanations (SHAP) is implemented by inputting the trained XGBregressor ML model. The following 3D printing process parameters, layer thickness, wall thickness, infill density, infill pattern, nozzle temperature, bed temperature, print speed, material, fan speed are considered to predict the tensile strength, roughness and elongation. SHAP explanations clarify process parameters’ proportional and cumulative effects on anticipated qualities. The XGBoost model achieved a mean squared error (MSE) of 0.591 and root mean squared error (RMSE) of 0.769. SHAP visualization plots are presented to understand the patterns of interaction between the most influential process parameters. The plots revealed that layer height positively correlates with roughness, while nozzle temperature is the most influential factor for tensile strength. Infill density is key for elongation, with higher infill leading to higher predicted elongation. This knowledge can be used to prioritize parameter optimization and control for achieving desired material properties, ultimately leading to more reliable and consistent 3D printing processes.
