Automatic Design System With Generative Adversarial Network and Convolutional Neural Network for Optimization Design of Interior Permanent Magnet Synchronous Motor

Shimizu, Yuki; Morimoto, Shigeo; Sanada, Masayuki; Inoue, Yoshio

doi:10.1109/tec.2022.3208129

Cited by 28 publications

(9 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Over the recent years, machine learning techniques—and in particular NNs—have been used to address various aspects related to magnetic losses in electrical engineering and in power electronics. NNs of different sizes and complexities have been trained to provide a hysteresis model, 11,12 a soft magnetic composite core model, 13 or even full machine models, 14 with various assumptions and geometric simplifications. Mušeljić et al 15 uses a feed‐forward NN to identify the parameters of the energy‐based hysteresis model 16,17 …”

Section: Introductionmentioning

confidence: 99%

Neural network‐based simulation of fields and losses in electrical machines with ferromagnetic laminated cores

Purnode,

Henrotte,

Louppe

et al. 2024

Int J Numerical Modelling

View full text Add to dashboard Cite

Due to the distribution of eddy currents inside ferromagnetic laminations, the accurate modeling of magnetic fields and losses in the laminated cores of electrical machines requires resolving individual laminations with a fine 3D discretization. This yields finite element models so huge and costly that they are unusable in daily industrial R&D. In consequence, hysteresis and eddy currents in laminations are often simply disregarded in the modeling: the laminated core is assumed to be made of a reversible (non lossy) saturable material, and magnetic losses are evaluated a posteriori, by means of Steinmetz–Bertotti like empirical formulas. However, in a context where industry is struggling to minutely assess the impact of magnetic losses on their devices, this simplified approach is more and more regarded as inaccurate and unsatisfactory. This article proposes a solution to this issue, based on homogenization and on detailed mesoscopic simulations of eddy currents and hysteresis inside the laminations. The proposed approach results in a close‐to‐conventional 2D magnetic vector potential finite element model, but equipped with an irreversible parametric material law to represent the ferromagnetic stack. In each finite element, the parameters of the law are obtained from a neural network trained to best fit the detailed mesoscopic simulations of the laminations subjected to the same local magnetic field. This way, all aspects of the irreversible ferromagnetic response are appropriately accounted for in the finite element simulation, but at a computational cost drastically reduced with regard to a brute force 3D calculation, and comparable to that of conventional 2D finite element simulations.

show abstract

Section: Introductionmentioning

confidence: 99%

Neural network‐based simulation of fields and losses in electrical machines with ferromagnetic laminated cores

Purnode,

Henrotte,

Louppe

et al. 2024

Int J Numerical Modelling

View full text Add to dashboard Cite

show abstract

“…Torque and area are used as conditional labels; the trained GAN model outputs new shapes that indicates the required performance values by inputting the requirements as labels. In the literature, [22] and [23] used GAN for designing motors. In these studies, the input data for the GAN model was an image.…”

Section: Introductionmentioning

confidence: 99%

“…In these studies, the input data for the GAN model was an image. Further, [22] defines the shape by setting the design variables (e.g., magnet width and position) and creating the CAD data. Since all data are represented by a set of parameters, the GAN output remains limited.…”

Section: Introductionmentioning

confidence: 99%

Automotive motor rotor design synthesis using cWGAN-gp with distortion penalty

Kato,

Suzuki,

Kondo

et al. 2024

Preprint

View full text Add to dashboard Cite

This study proposes a motor-rotor-shape generation method based on a conditional Wasserstein generative adversarial networks (GAN) with gradient penalty (cWGAN-gp) to perform the initial design of the automotive motor efficiently. We introduced a distortion degree as a regularization in the training process to generate smooth, intuitively understandable, and easy to manufacture shapes. Rotor shapes generated by cWGAN-gp exhibited characteristics similar to a combination of multiple training data with performances close to that of the input label data. Furthermore, smooth shapes were generated by appropriately setting the influence of the distortion degree. The motor shapes could be generated accurately when a sufficient amount of training data existed. These results suggest that training the model with accumulated motor data allows for the rapid generation of new shapes based on input performance criteria in a short period of time.

show abstract

“…The control system's effectiveness is greatly impacted by the settings selected. Numerous academics have conducted thorough investigations on parameter tuning, and artificial intelligence algorithms offer a promising approach to address parameter optimization challenges in intricate nonlinear systems [26][27][28][29]. Ref.…”

Section: Introductionmentioning

confidence: 99%

Research on the Control System of Bearingless Induction Motor Based on Improved Active Disturbance Rejection Control

Wang,

Yang,

Sun

et al. 2024

PIER C

View full text Add to dashboard Cite

To enhance the reaction speed, suspension performance, and anti-interference ability of a Bearingless Induction Motor (BIM) operation control system, an improved Active Disturbance Rejection Control (ADRC) technique is proposed. Firstly, the ADRC in the suspension system and the ADRC in the torque system are designed, respectively, using the BIM's mathematical model as the basis. Furthermore, the error integral signal is incorporated into the nonlinear state error feedback control law of the standard ADRC controller. Subsequently, a novel optimal control function is formulated using the fitting method, which is based on the original fal function. This approach effectively mitigates the impact of output signal fluctuations at the inflection point of the fal function. Simultaneously, the RBF neural network technique is employed to autonomously adjust the control parameters of the extended state observer, therefore enhancing the system's observation capability. Ultimately, the classic ADRC control strategy and the IADRC strategy are compared through simulation and experimentation. Simulations and experimental findings demonstrate that the suggested control method enhances the BIM control system's response time and resilience to external disturbance. Additionally, it enhances the levitation performance of the BIM system.

show abstract

Automatic Design System With Generative Adversarial Network and Convolutional Neural Network for Optimization Design of Interior Permanent Magnet Synchronous Motor

Cited by 28 publications

References 23 publications

Neural network‐based simulation of fields and losses in electrical machines with ferromagnetic laminated cores

Neural network‐based simulation of fields and losses in electrical machines with ferromagnetic laminated cores

Automotive motor rotor design synthesis using cWGAN-gp with distortion penalty

Research on the Control System of Bearingless Induction Motor Based on Improved Active Disturbance Rejection Control

Contact Info

Product

Resources

About