基于NSGA-Ⅱ和BP神经网络的螺旋板式换热器多目标优化  

Multi-objective Optimization of Spiral Plate Heat Exchanger Based on NSGA-Ⅱ and BP Neural Network

作　　者：郭俊飞 卿德藩[1] 倪浪 白建军 向立平 Guo Junfei;Qing Defan;Ni Lang;Bai Jianjun;Xiang Liping(College of Mechanical Engineering,University of South China,Hengyang,421001;School of Nuclear Science and Technology,University of South China,Hengyang,421001)

机构地区：[1]南华大学机械工程学院,衡阳421001 [2]南华大学核科学技术学院,衡阳421001

出　　处：《制冷与空调（四川）》2024年第2期161-167,共7页Refrigeration and Air Conditioning

基　　金：湖南省教育厅科学研究重点项目(20A429)。

摘　　要：提出了一种带扰流片的螺旋板式换热器结构参数多目标优化方案,该方案主要采用快速非支配排序遗传算法NSGA-Ⅱ、BP神经网络以及CFD正交试验,在给定的范围内寻找最佳的扰流片结构参数。设置16组正交仿真试验模型,并根据试验样本数据得到扰流片结构参数和优化变量之间的非线性映射关系。以对流换热系数和压降为目标函数,使用NSGA-Ⅱ算法对优化后的BP神经网络进行极值寻优,得出优化目标下的Pareto解集,最后采用TOPSIS评价方法得到扰流片结构多目标优化的最终结果。仿真结果表明,优化后的结构同多目标优化算法的预测值误差都在15%以内,对流换热系数h较传统的螺旋板式换热器提高31.4%,说明该优化方法是可行的,为换热器结构优化和工程应用提供一定的参考。In this paper,a multi-objective optimization scheme for structural parameters of spiral plate heat exchanger with spoilers is proposed.This project mainly uses NSGA-Ⅱ,BP neural network and CFD orthogonal test to find the optimal structural parameters of spoilers in the given environment.First,the nonlinear mapping relationship between spoiler structure parameters and optimization variables can be obtained according to the analysis on the data from 16 groups of models set up for orthogonal simulating test.Then,take convective heat transfer coefficient and pressure drop as objective function,using NSGA-Ⅱalgorithm to optimize the extremum of the optimized BP neural network to obtain the Pareto solution set under the optimization objective.Finally,the multi-objective optimization result about the spoiler structure can be acquired through the utilization of TOPSIS evaluation method.The simulation results show that the predicted value error of the optimized structure and the multi-objective optimization algorithm is less than 15%,and the convective heat transfer coefficient h is increased by 31.4%compared with the traditional spiral plate heat exchanger.It indicates that the optimization method suggested in this paper is well practicable for the heat exchangers’structure improvement and engineering application to providea worthwhile reference.

关 键 词：遗传算法 BP神经网络 正交试验 换热器 多目标优化 

分 类 号：TH123[机械工程—机械设计及理论]