电工技术学报2024，Vol.39Issue(10)：3180-3191,12.DOI:10.19595/j.cnki.1000-6753.tces.230429
基于多岛遗传算法与响应面法的横向磁通感应加热装置参数优化设计
Optimization Design of Parameters for Transverse Flux Induction Heating Device Based on Multi-Island Genetic Algorithm and Response Surface Method
摘要
Abstract
Transverse flux induction heating(TFIH)devices are widely used in the heat treatment process of flat workpieces,such as metal strips.The heating results affect the application performance of metal strips.Ideal average temperature and uniform temperature distribution of the strip at the heater outlet are desired when the heating state of the strip is stable.The standard tempering temperature of 45#steel is 600℃.This paper divides the optimization parameters into structural and power supply parameters to meet heat treatment requirements.Different methods are used to reduce optimization difficulties while maintaining optimization accuracy. The structure of the TFIH device is determined by seven parameters.The air gap between the magnetic pole and the strip should be as small as possible to meet the processing technology.A global sensitivity analysis(GSA)based on the Morris method ranks the sensitivity of the relative non-uniformity to the six coil structural parameters.The top four structural parameters with significant effects are screened out,ensuring optimization accuracy while reducing calculation time. Due to the strong robustness,the radial basis function(RBF)neural network prediction model replaces the finite element calculation model to estimate nonlinear functions.The optimal Latin hypercube design(OLHD)samples 100 times,and the sampling results are used as the training sample points of the prediction model.After testing,the RBF neural network model has high prediction accuracy within the variation range of the input parameters.The multi-island genetic algorithm(MIGA)optimizes the screened structural parameters globally.The results show that the relative non-uniformity is reduced from 2.88%to 2.38%after optimization,effectively improving the temperature distribution uniformity. Based on the optimized structural parameters,the response surface method(RSM)is used to optimize the current and frequency.Consequently,the average temperature is close to the target value,and the relative non-uniformity is maintained at a low level.The relative non-uniformity and the average temperature expressions for power supply parameters are fitted separately using a second-order polynomial.Both models are tested by the variance(ANOVA)analysis.The multi-objective optimization is then performed using the response optimizer in Design-Expert.The results show that after the optimization of structural parameters and power supply parameters,when the strip heating state reaches stability,the relative non-uniformity of the strip surface temperature distribution at the heater outlet is 2.36%,and the average temperature is 600.06℃,which can meet the tempering requirements of 45#steel strip.关键词
横向磁通感应加热/全局灵敏度分析/径向基函数神经网络模型/多岛遗传算法响应面法Key words
Transverse flux induction heating/global sensitivity analysis/radial basis function neural network model/multi-island genetic algorithm/response surface method分类
矿业与冶金引用本文复制引用
刘志赢,汪友华,刘成成,彭江湃,宋华宾..基于多岛遗传算法与响应面法的横向磁通感应加热装置参数优化设计[J].电工技术学报,2024,39(10):3180-3191,12.基金项目
河北省省级科技计划资助项目(215676146H,225676163GH). （215676146H,225676163GH）
