This study optimizes 3D print orientation to minimize support material, printing time, and surface roughness using non-dominated sorting genetic algorithm II (NSGA-II). Traditional NSGA-II can stagnate due to static parameters; thus, integration with Q-learning dynamically adjusts these parameters based on rewards. Q-learning, a variant of reinforcement learning (RL), initially promotes population diversity for broad exploration and later exploits optimal solutions near the Pareto front. Results show that the hybrid approach significantly enhances Pareto-front quality, improving efficiency by reducing support material (2.1%), printing time (3.8%), and surface roughness (1.9%). Validation confirms practical applicability of generated solutions for fused deposition modeling (FDM).