Abstract: In order to improve the comprehensive performance of submersibles, the discipline decomposition of submersibles was carried out based on the idea of multidisciplinary design optimization. The design parameters and data transfer relationships in the overall design model were determined. The secondary developments of ABAQUS software and STAR-CCM+software were carried out, and the structure discipline and boat shape discipline parametric analysis flows were designed, the main body parametric analyses were realized as well. A fourth order response surface model was used to establish the approximate model of boat shape discipline and structure discipline. The propulsion parameters were fitted, and the approximate model of the propulsion discipline was established. The mechanical electrical equipment discipline was treated as quantitation. Relationship among energy discipline, propulsion discipline and mechanical electrical equipment discipline was determined. Also, the analysis model of energy discipline was established. Based on the discipline analysis model, the system levels were analyzed. Then the sensitivity analyses of design variables were carried out. A multidisciplinary optimization model of the submersibles was established based on multidisciplinary feasible method, and the second generation of non dominated sorting genetic algorithm (NSGA-Ⅱ) was used to solve the problem. Results show that the optimization effectiveness of the selected scheme is obvious and the comprehensive performance of submersibles is improved.

Key words: submersible, discipline decomposition, parameterization, approximate model, optimization model

摘要： 为提高潜水器的综合性能，基于多学科设计优化思想，进行了一型潜水器学科分解，确定了设计参数及总体设计模型的数据传递关系。进行了ABAQUS软件和STAR-CCM+软件的二次开发，设计了结构学科和艇型学科参数化分析流程，实现了主体参数化分析。采用四阶响应面模型建立了艇型学科、结构学科近似模型，对推进器参数进行了拟合，建立了推进学科近似模型，对机电设备学科进行了定量处理，确定了能源学科与推进学科、机电设备学科之间的关系，建立了能源学科分析模型。基于学科分析模型，对系统层进行了分析，同时进行了设计变量灵敏度分析，建立了基于多学科可行方法的潜水器多学科优化模型。利用第二代非支配排序遗传算法（NSGA-Ⅱ）进行了求解，在Pareto解集中选择部分方案与初始方案对比，结果表明：所选择的方案优化效果明显，提升了潜水器的综合性能。

关键词: 潜水器, 学科分解, 参数化, 近似模型, 优化模型