基于神经网络与遗传算法节能扰流子优化设计  

Optimization of Energy Saving Turbolator Based on BP Neural Network and GA Algorithm

作　　者：宋乾斌 陆晓峰[1] 卢培业 朱晓磊[1] SONG Qian - bin;LU Xiao - feng;LU Pei - ye;ZHU Xiao - lei(College of Mechanical and Power Engineering,Nanjing Tech University,Nanjing Jiangsu 211816,China)

机构地区：[1]南京工业大学机械与动力工程学院,南京211816

出　　处：《计算机仿真》2018年第8期255-260,共6页Computer Simulation

基　　金：江苏省"六大人才高峰"项目(2015-ZBZZ-013)

摘　　要：在弯管前安装扰流子,可以减小弯管处二次流强度,降低能量损失,并运用CFD软件对不同参数下的扰流子节能效果数值计算。以L9(33)正交试验以及4组补充试验作为BP神经网络的训练样本,建立在5种雷诺数下扰流子节能效率与扰流子叶片转角、叶片长度、安装距离3个结构参数的非线性映射关系;扰流子节能效率最大值作为目标函数,再结合遗传算法进行结构参数优化。最终得到在不同雷诺数下扰流子叶片转角、叶片长度、安装距离的最佳组合形式,并利用有限元方法对结果验证。结果表明,这种优化方案具有可行性;合适的结构参数的扰流子具有良好的节能效果。A turbolator is installed in front of bend for diminishing the strength of secondary flow and reducing the energy loss of elbow. The energy - saving effect of turbolator with different structure parameters were simulated with CFD software. The L9 （33） -orthogonal experiments and fore complementary experiments were chosen as the trained samples of Back Propagation Neural Network. The structure parameters of turboaltor were rotation angle, vane length, and mounting distance. The nonlinear mapping relation between the structures and energy - saving efficiency waw built up. In order to get the maximum energy efficiency of turbolator, the structure parameters were optimized with BP- GA algorithm. Under different Reynolds number, the fittest turbolator structure parameters of rotation angle, vane length and mounting distance were found. And the results were also verified through the finite element method. The results show that the optimization scheme was feasible, and this turbolator with suitable structure parameters has a good energy saving effect.

关 键 词：扰流子 节能 神经网络 遗传算法 湍流 

分 类 号：TK018[动力工程及工程热物理]