Multi-Level Modeling Methodology for Optimal Design of Electric Machines Based on Multi-Disciplinary Design Optimization

Abstract

The transportation sector is undergoing electrification to gain advantages such as lighter weight, improved reliability, and enhanced efficiency. As contributors to the safety of embedded critical functions in electrified systems, better sizing of electric machines in vehicles is required to reduce the cost, volume, and weight. Although the designs of machines are widely investigated, existing studies are mostly complicated and application-specific. To satisfy the multi-level design requirements of power systems, this study aims to develop an efficient modeling method of electric machines with a background of aircraft applications. A variable-speed variable-frequency (VSVF) electrically excited synchronous generator is selected as a case study to illustrate the modular multi-physics modeling process, in which weight and power loss are the major optimization goals. In addition, multi-disciplinary design optimization (MDO) methods are introduced to facilitate the optimal variable selection and simplified model establishment, which can be used for the system-level overall design. Several cases with industrial data are analyzed to demonstrate the effectiveness and superior performance of the modeling method. The results show that the proposed practices provide designers with accurate, fast, and systematic means to develop models for the efficient design of aircraft power systems.

Recent Developments in Electrical Machine Design for the Electrification of Industrial and Transportation Systems

“…With the present level of resources, these numerical tools can be applied to computationally intensive optimization methods. Evolutionary algorithms are a popular example of such methods, such as particle swarm optimization (PSO) [10] and the non-dominated sorting genetic algorithm (NSGA) [11,12]. This coupling leads to designs that push the limits of the constraints imposed by the considered topology, current materials, and technologies, thus uncovering new development directions.…”

“…In [11], the authors propose a multi-disciplinary design optimization (MDO) methodology based on analytical models for the design of electrical machines appropriated for incorporation into a system-level design process. The MDO method allows considering the couplings between the several components of machine design considering structural, magnetic, and thermal features, Figure 3a.…”

“…In this case, work is invested in creating the model and checking its accuracy to reduce the computational time required for the optimization. The focus of [11] is on an aircraft application, specifically the auxiliary power unit generator in a Boeing 787, which is made of a variable-speed variable-frequency salientpole synchronous generator, Figure 3b. Compared to a baseline solution made by a PMSG, replacing the electrical excitation modules (silicon steel pole modules and excitation winding modules) with NdFeB magnet poles, the presented MDO methodology and its simplified surrogate generator model produce agreeable results with a 1.06% deviation on the estimated machine weight and −2.03% on power loss, with lower complexity, requiring only six inputs.…”

Enhancing design automation for components of electric machines: a systematic approach

“…Thus, both system-centric and component-centric design approaches are constrained by their inherent limitations. Although some authors, such as Dai et al (2019), adopt strategies that consider both component and system design aspects, the review by Umland et al (2023) highlights that multilevel approaches are often overlooked in electric motor design, thus missing opportunities for parallel optimization processes and indicating a need for further methodological support.…”

Electrical machines surrogate‐based design optimization based on novel waveform targeting strategy with improvement of the computational efficiency

“…Accordingly, one of the noteworthy issues that causes considerable difficulties in the surrogate modelling process is the high-dimensional design spaces with numerous input variables. Therefore, advanced approaches such as multi-level optimization strategies have been adopted to tackle this drawback [17][18][19]. Furthermore, a harsh and disordered behaviour of output functions against input variables can considerably affect the accuracy of the models.…”