<p>Laser powder bed fusion (LPBF) presents significant challenges due to the complexities of multi-scale, multi-physics coupling and the need for multi-parameter control. These challenges make multi-objective optimization for quality improvement particularly difficult. To overcome these obstacles, a data-driven intelligent modeling and multi-objective optimization framework is proposed for LPBF of 316L powders. It consists of three interrelated and logistically coordinated components: data-driven mechanistic modeling, optimization-driven data generation, and property-driven multi-objective optimization. By combining powder-scale multi-physics simulation with an AI-assisted data generation model, the framework establishes the relationships between process parameters and melt pool characteristics, grounded in a comprehensive physical understanding. This approach significantly enhances efficiency, accelerating the process by several orders of magnitude. Moreover, as a kind of correlation elimination and dimension reduction method, the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) model integrated with Mahalanobis distance is capable of solving multiple-objective optimization problems rapidly, which has been well-validated through experimental measurements. The optimization procedure is performed using a database of 74 cases, incorporating data from post-process measurements, numerical simulations, and artificial neural network (ANN) predictions. Then the relative closeness of each alternative is calculated for the final rank ordering. Ultimately, case 27 is identified as the optimal solution, meeting the overall objective target of the optimization which includes three predefined beneficial sub-objectives: melt pool height, width/depth ratio, and track morphology. This innovative approach offers informative guidance for practical processes, significantly enhancing efficiency while focusing on multi-objective optimization.</p>