.png){kind=icon}
.png){kind=icon}
Return
Multi-objective optimization based on hyperparameter random forest regression for linear motor design
.png){kind=icon}
PRE
AI
DOI：10.1007/s13042-022-01573-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Existing method of multi-objective optimization for linear motors provides poor consideration for model robustness, thus, the models of linear motor which are implemented to multi-objective optimization may not be optimal. A multi-objective optimization design combined Random Forest after hyperparameter optimization and non-dominated sorting genetic algorithm-II (NSGA-II) for Double-sided Linear Flux Switching Permanent Magnet motor (DLFSPMs) is proposed. The average thrust and the thrust ripple which generated by operation of DLFSPMs are selected as objectives. A machine learning algorithm, Random Forest (RF), is introduced to establish the regression models between structural parameters and performances of DLFSPMs. Furthermore, in order to improve the stability and accuracy of the regression models, a hyperparameter optimization, which is called Bayesian Optimization and HyperBand (BOHB), is proposed to search for the best hyperparametric configuration to obtain predicted performance of DLFSPMs. Moreover, the proposed BOHB-RF model is compared with the Bayesian optimization-RF (BO-RF) model and the HyperBand-RF (HB-RF) model to verify the advantages of BOHB-RF model. Then, NSGA-II is adopted to design for multi-objective optimization of DLFSPMs to calculate Pareto front of DLFSPMs performances based on BOHB-RF models. Finally, the results of finite element analysis (FEA) prove the effectiveness and feasibility for proposed modeling and optimizing method.
Keywords:
Linear motor
Multi-objective optimization
Random forest
Bayesian optimization and HyperBand
Non-dominated sorting genetic algorithm-II
Journal
IF:
2.7
Papers:
3.1K
Citations:
5.6K