Abstract: Existing multi-objective optimization methods for machining often underuse process-parameter sensitivity and mechanism information, leading to local optima and poor solution diversity. Therefore, based on the sensitivity-mechanism integrated multi-objective particle swarm optimization (SMG-MOPSO) algorithm, regression models and empirical formulas were proposed. Firstly, a prediction model for surface roughness (Ra), cutting force (F_c), and material removal rate (MRR) was constructed. Subsequently, sensitivity functions together with main-effect and interaction analyses were employed to reveal the influence patterns of process parameters on the objectives. Based on these insights, three mechanisms were incorporated into MOPSO, adaptive step size guided by sensitivity values and mechanism trends, velocity updating that integrated sensitivity guidance with mechanism-based correction, and solution-set maintenance with a mechanism-consistency criterion to enhance solution set effectiveness. Milling experiments showed that the proposed method increased MRR by about 24.40% while ensuring Ra and F_c remain within constraints, thereby demonstrating its effectiveness and engineering applicability.