A General SVM-Based Multi-Objective Optimization Methodology for Axial Flux Motor Design: YASA Motor of an Electric Vehicle as a Case Study

Abstract

Axial flux motor design is normally depended on a designer's experience to adjust design parameters, which is vague and complex, for example, torque density and torque ripple are two key factors of a motor to restrain its development, since torque density dominates a motor's volume and weight, while torque ripple determines its stability. Therefore, a general optimizations methodology is required in its design process. To realize this purpose, this paper proposes a general multi-objective optimization methodology for practical motor design. In detail, this methodology is based on Support Vector Machine-Chaotic Cultural Differential Evolution(SVM-CCDE) algorithm for both maximizing the torque density and minimizing the torque ripple, and Yokeless And Segmented Armature(YASA) motor of an electric vehicle(EV) considering practical constraints is presented as a typical example since it is a special topology of axial flux motor(AFM). A comparative analysis is presented in the paper to demonstrate the proposed method's advanced features. Finally, the effectiveness of the optimization method is verified by finite element analysis (FEA) via Ansys software, the results well agree the analyses and further validate the proposed method. The proposed method is a potential feasible solution to improve the EV's motor design in the coming future.INDEX TERMS Support vector machines-chaotic culture differential evolution algorithm, axial flux machine, multi-objective optimization.

Few‐sample multi‐objective optimisation of a double‐sided tubular machine with hybrid segmented permanent magnet

“…Popular regression model such as extreme learning machine (ELM) and K‐nearest neighbour (KNN) also requires hundreds of samples to complete training [20, 26]. Support vector regression (SVR) is available for modelling with a limited number of samples [27], however has multiple parameters to be tuned concurrently. Different from models aforementioned, a neural network, GRNN, can form a low‐error regression surface from only a small sample set with merely one parameter to be tuned.…”

Multi-Objective Optimization Design of Permanent Magnet Torque Motor

“…Suppression of torque ripple is an important content in the design of high-precision permanent magnet motor. The motor torque ripple is shown as [25]:…”

“…In the paper, SVM is used to fit the data and the relationship between the design parameters and the optimization objective can be obtained. The SVM formula is expressed as [25]:…”