基于序贯Kriging模型的潜器型线优化设计

潜器型线优化设计是一个多目标优化问题,在型线设计过程中,阻力性能与包络体积的要求是相互冲突的。为了解决计算流体力学软件如Fluent在进行潜器的外形优化设计时效率低下问题,采用Kriging模型代替仿真模型进行潜器外形设计的策略,其基本思想是：选取设计变量和样本点,利用ICEM软件建立参数化的水动力分析模型,用Fluent软件计算得到样本点的阻力响应值,建立反映设计变量与响应之间关系的Kriging模型,将阻力和体积作为潜器外形优化的两个目标,利用多目标遗传算法求出Pareto最优解。由于采样策略对Kriging模型精度影响很大,本文提出了一种新的序贯采样方法命名为加权累积误差方法,来选取样本点以提高Kriging模型精度。结果表明提出的序贯Kriging建模技术能极大提高潜器型线优化设计效率,同时保证设计精度。

Shape optimization design of submersible based on sequential Kriging model

Shape optimization design of submersible is a multi-objective optimization problem, the requirements of resistance and envelope volume are in conflict in the design process. To solve the problem that computational fluid dynamics software, such as Fluent, is inefficient in shape optimization design of submersible, Kriging model is used to replace the computational simulation model. The basic idea of this strategy is as follows: Design variables and sample points are selected and the parameterized grid model is set up in ICEM. Fluent is used to calculate the resistance. Kriging model which approximates the relationship between design variables and resistance is established. The resistance and envelope volume are set as two objectives and Pareto front is obtained by using multi-objective genetic algorithm (MOGA). Sampling strategy is an important factor that affects the accuracy of a given metamodel. A novel sequential sampling strategy, named weighted accumulative error sampling (WAES) approach, is proposed to improve sampling quality for metamodeling. The results indicate that the proposed sequential Kriging modeling approach can greatly improve the efficiency of the shape optimization design of submersible, guarantee the design precision and provide a reference for solving similar shape design problems.