基于神经网络和遗传算法的温差发电器优化设计  

Optimization design of thermoelectric generator based on neural network and genetic algorithm

作　　者：何嘉华 周宏甫[1] 刘二辉[1] 张征[1] HE Jia-hua;ZHOU Hong-fu;LIU Er-hui;ZHANG Zheng(School of Mechanical and Automobile Engineering,South China University of Technology,Guangzhou 510641)

机构地区：[1]华南理工大学机械与汽车工程学院,广东广州510641

出　　处：《机械设计》2018年第9期31-36,共6页Journal of Machine Design

基　　金：国家自然科学基金资助项目(51375168)

摘　　要：提出了将温差发电器对内燃机排气背压的影响纳入温差发电器的优化设计过程的观点,设计了一套新的温差发电器优化方案。以发电器尺寸参数为设计变量,以排气背压、质量作为约束条件,以发电片温差为目标进行优化设计。利用中心复合设计法选取试验点,对试验点进行CFD仿真,采用高预测精度的改进BP神经网络拟合设计变量与目标函数间的关系,再利用遗传优化算法在设计空间寻找最佳设计点。优化后消除了发电器对排气背压的影响,温差提高了8.8%,质量降低了6.7%。For the optimization design of thermoelectric generator, the influence of thermoelectric generator exerted on the exhaust back pressure of the internal combustion engine is taken into consideration. A new scheme for the optimization design of thermoelectric generator is formulated. The dimension parameters are taken as design variables; the exhaust back pressure and the mass are taken as constraints; the temperature difference of power module is taken as the objective of optimization design. By means of the central composite design and the CFD simulation, the design points are selected and tested. The improved BP neural network characterized with high-prediction accuracy is alopted to fit the relation between the design variables and the objective function. By means of genetic algorithm, the best design points in the design space are identified. Based on the optimization, the influence of thermoelectric generator exerted on the exhaust back pressure is eliminated. As a result, the temperature difference increases by 8.8%, and the mass drops by 6.7%.

关 键 词：温差发电器 神经网络 中心复合设计法 遗传算法 

分 类 号：TK124[动力工程及工程热物理—工程热物理]