Disruptive innovations in electrical machine design optimization are observed in this paper, motivated by emerging trends. Improvements in mathematics and computer science enable more detailed optimization scenarios that cover evermore aspects of physics. In the past, electrical machine design was equivalent to investigating the electromagnetic performance. Nowadays, thermal, rotor dynamics, power electronics, and control aspects are included. The material and engineering science have introduced new dimensions on the optimization process and impact of manufacturing, and unavoidable tolerances should be considered. Consequently, multifaceted scenarios are analyzed and improvements in numerous fields take effect. This paper is a reference for both academics and practicing engineers regarding recent developments and future trends. It comprises the definition of optimization scenarios regarding geometry specification and goal setting. Moreover, a materials-based perspective and techniques for solving optimization problems are included. Finally, a collection of examples from the literature is presented and two particular scenarios are illustrated in detail.
17 June 2020 in IEEE Transactions on Energy Conversion
This paper presents a new method for multi-objective robust design optimization of electrical machines and provides a detailed comparison with so far introduced techniques. First, two robust design approaches, worst-case design and design for six-sigma, are compared with the conventional deterministic approach for multi-objective optimization. Through a case study on a permanent magnet motor, it is found that the reliabilities of motors produced based on robust designs are 100% under the investigated constraints, while the reliabilities of deterministic designs can be lower than 30%. A major disadvantage of robust optimization is the huge computation cost, especially for high-dimensional problems. To attempt this problem, a new multi-objective sequential optimization method (MSOM) with an orthogonal design technique and hypervolume indicator (as a measure of convergence) is proposed for both deterministic and robust design optimization of electrical machines. Through another case study, it is found that the new MSOM can improve motor performance and greatly reduce the computational cost. For the robust optimization, the number of required finite element simulations can be reduced by more than 40%, compared with that required by the conventional approach. The proposed method can be applied to many-objective (robust) design optimization of electrical machines.
This paper reviews the recent developments of design optimization methods for electromagnetic devices, with a focus on machine learning methods. First, the recent advances in multi-objective, multidisciplinary, multilevel, topology, fuzzy, and robust design optimization of electromagnetic devices are overviewed. Second, a review is presented to the performance prediction and design optimization of electromagnetic devices based on the machine learning algorithms, including artificial neural network, support vector machine, extreme learning machine, random forest, and deep learning. Last, to meet modern requirements of high manufacturing/production quality and lifetime reliability, several promising topics, including the application of cloud services and digital twin, are discussed as future directions for design optimization of electromagnetic devices.
15 July 2020 in IEEE Transactions on Energy Conversion
In this paper, a novel system-level robust design optimization method is presented to improve the performance of switched reluctance motor (SRM) drive systems under multiple operating conditions. Based on typical driving cycles of electric vehicles (EVs), five typical driving modes of the SRM are determined. The optimization objectives in each driving mode are established. The significant parameters of the motor and controller of each driving mode are selected as the optimization variables by using the sensitivity analysis. In order to simplify the optimization process, correlation analysis is performed to determine the coherence of the objective functions of all driving modes. Then, a sequential Taguchi method is applied to find an optimal design which is less sensitive to the noise factors. To verify the effectiveness of the proposed method, an SRM drive system applied in EVs with a 12/10 SRM and angle position control method is investigated. It is found that the proposed method can significantly reduce the torque ripple and improve the comprehensive performance. Finally, a 12/10 SRM is prototyped and tested to validate the simulation results.
4 August 2022 in Renewable and Sustainable Energy Reviews
Multi-objective optimization algorithms are becoming ever more popular in the field of electrical machine design as they provide engineers with an automated way of efficiently exploring huge design spaces when searching for machines that are simultaneously highly competitive regarding several objectives, such as efficiency, material costs, torque ripple, and others. Apart from exhibiting these good target characteristics, a good design should also be robust, i.e., it should not be very sensitive to slight changes in its design parameters as this would either seriously impact production costs or make the physical machine behave differently than its (optimized) computer simulation model. This paper is focused on describing how global surrogate models (i.e., nonlinear regression models), that are created in order to reduce the dependence on finite-element (FE) simulations during the multi-objective optimization run, can be easily reused to perform very fast local and global tolerance/sensitivity analyses of generated designs. While obtained in a fraction of the time required by the complementary FE-based approach, the surrogate-based sensitivity estimates are able to provide accurate and valuable information regarding the robustness of electrical machine designs. Ultimately, by integrating robustness-related information with Pareto front projections, we aim to provide engineers with much clearer pictures of the specific problem-related tradeoffs discovered by the automated design optimization procedure.