利用CFD软件STAR-CCM+对螺旋缠绕式湍流促进器的8种设计参数进行流场数值计算，以壁面剪应力和轴向压降为指标，比较不同参数的湍流促进器强化效果。建立了基于遗传算法优化的BP神经网络，使用神经网络对不同的设计参数进行流场效果预测，网络拟合精确度达到了0.99 718。以最小化轴向压降最大化平均壁面剪应力为目标，利用NSGA-Ⅱ算法进行多目标优化，计算出Pareto最优支配前沿，在最优支配前沿中寻找到合适的设计点，并对预测出的优化参数进行建模并模拟验证。结果分析表明，与初始设计参数相比，在中心杆直径选取为10.8 mm、螺距为6.5 mm时，流场的轴向压降增加了1.7%，平均壁面剪应力提高了9.7%。

In order to design a turbulence accelerator with appropriate parameters,8 design parameters are simulated by using CFD software STAR-CCM+ in consideration of wall penetration.The strengthening effects of different parameters are compared with shear stress and axial pressure drop as indexes.BP neural network based on genetic algorithm optimization is established and the flow field effect is predicted by using neural network for different design parameters.Nsga-ⅱ algorithm is used to do multi-objective optimization to generate Pareto optimal frontier by taking minimum axial pressure drop and maximum average wall shear stress as the objectives.Appropriate design points are found among the optimal frontier and the predicted optimization parameters are modeled and simulated.The results show that the axial pressure drop of the optimized turbulent flow accelerator increases by 1.7% compared with the initial design parameters,but the average wall shear stress increases by 9.7% when the diameter of center rod is 10.8 mm and the pitch is 6.5 mm.

郗元(1987-),男,博士,工程师,研究方向为流场分析及热管理,xiyuan@dlut.edu.cn