双流道泵蜗壳多目标多学科设计优化

为实现同时提高双流道泵水力性能和结构性能的研究目标,采用均匀试验设计、高精度流固耦合计算、人工神经网络建模以及多目标遗传寻优相结合的方法,选取蜗壳的进口直径、进口宽度、隔舌安放角和扩散段长度为优化变量,利用均匀试验设计方法设计了50组试验,通过流固耦合计算方法得到双流道泵设计工况点的效率以及蜗壳段的最大应力,并以此为优化目标,通过BP人工神经网络训练建立近似函数,设计了离心泵专用的多目标多学科遗传寻优策略对目标函数进行寻优,得到了蜗壳几何参数组合的Pareto前沿。结果表明:相对于原始方案,优化方案的扩压效果明显改善,并且减少了隔舌处及扩散段的回流现象;优化方案的效率得到提升;优化方案的最大应力和最大应力点的平均振动速度显著降低。

Multi-objective and Multidisciplinary Optimization of Double-channel Pump

In order to improve the hydraulic and structural performances of the double-channel pump, the uniform test design, fluid-structure interaction calculation, combination of artificial neural network and multi-objective genetic optimization method were applied. The inlet diameter, inlet width, tongue angle and diffusion length were selected as optimization variables, which were then used to produce 50 tests by using the method of uniform test design. The efficiency and maximum stress, attained through fluid-structure interaction in a volute at the design condition, were chosen as the optimization goals and then were applied to get the approximate function by BP neural network training. Special multi-objective genetic optimization strategy was designed to solve the function and search the optimized results which are called the Pareto frontiers of geometry parameters of the volute. During the process of searching the Pareto frontiers, the efficiencies, heads and stresses were chosen as constrains to get the final two optimizations. The results showed that compared with the original case, pressure diffusion effects of the optimized cases were improved obviously, backflow phenomena was reduced, and the efficiencies of cases opt1 and opt2 were increased, while the maximum stresses and the mean vibration velocities of opt1 and opt2 were depressed. The results indicated that the proposed multi-objective and multidisciplinary optimization method can obviously improve the hydraulic and structural performances of double channel pump, meanwhile, it has some reference value for optimization researches on other turbomachineries.