• Layer thickness dominates processing time • Inter-layer bonding (weld) strength depends on temperature–speed balance • LightGBM and Random Forest accurately model weld strength and processing time • Surrogate-assisted NSGA-II enables multi-objective optimization of FDM parameters • NSGA-II–TOPSIS identifies balanced parameter sets for reliable FDM decision-making Fused Deposition Modeling (FDM) is widely used in additive manufacturing, yet weak inter-layer weld strength and long processing times continue to continue to limit its industrial applicability. This study presents a novel surrogate-assisted multi-objective optimization framework to maximize weld strength while minimizing processing time in PLA+-based FDM. Machine learning models—Random Forest (RF), Gradient Boosting, XGBoost, and LightGBM—were developed and trained to model the nonlinear relationships between extrusion temperature, extruder velocity, layer thickness, and process responses. LightGBM achieved the highest predictive accuracy for weld strength (test R 2 = 0.955, RMSE = 0.052 N/mm), while RF performed best for processing time prediction (test R 2 = 0.9996, RMSE = 5.18 s). These models were embedded as surrogates within a hybrid NSGA-II–TOPSIS framework to generate and rank Pareto-optimal solutions. The results revealed a clear trade-off between mechanical performance and production time: the highest weld strength of 0.84 N/mm was achieved at 225 °C and 0.1 mm layer thickness with a processing time of 1083 s, whereas the shortest processing time of 377 s resulted in a weld strength of 0.30 N/mm. A balanced solution selected using TOPSIS (225 °C, 50 mm/s, 0.2 mm) achieved 0.65 N/mm weld strength with a 49% reduction in processing time relative to the strength-optimal case. Experimental validation showed prediction errors below 3%. The proposed framework offers a practical and reliable decision-support tool for balancing strength and productivity in engineering-oriented FDM applications.
Chowdhury et al. (Fri,) studied this question.