Enhanced Tensile Strength Prediction for 3D-Printed Parts: Integrating Kookaburra Optimization with XGBoost Machine Learning

This study proposes a hybrid predictive framework that integrates the Kookaburra Optimization Algorithm with Extreme Gradient Boosting, referred to as KOA-XGB, for estimating the tensile strength of fused deposition modeled polylactic acid parts. A dataset consisting of 204 experimental samples, covering tensile strength values from 17.67 MPa to 57.65 MPa and five key printing parameters, was used for model development and validation. The proposed KOAXGB model achieved a coefficient of determination of 0.92 on the test set, with a mean absolute error of 2.15 MPa, representing a 23% improvement over standard XGBoost and a 35% improvement compared to random forest models. Approximately 95% of predicted values fell within ±3 MPa of experimental measurements, indicating high prediction consistency. Feature importance and SHAP analysis revealed that infill density and layer thickness were the dominant parameters, with mean SHAP values of 0.45 and 0.38, respectively. Increasing infill density from 20% to 80% resulted in an average tensile strength improvement of nearly 67%, while reducing layer thickness from 0.3 mm to 0.1 mm enhanced tensile strength by approximately 14%. The optimization process converged within 150 iterations, reducing test mean squared error to 5.32 and achieving convergence about 40% faster than conventional grid search methods. The findings demonstrate that the proposed KOA-XGB framework supports resource efficiency, responsible production, and sustainable manufacturing by reducing experimental effort by up to 45% and lowering material waste by approximately 20% to 25%, thereby contributing to datadriven decision making in additive manufacturing systems.