Double flux radial electrodynamic bearing with the radial air gap between permanent magnet rings and its optimization using artificial neural network for bearing characteristics

Document Type

Article

Publication Title

Engineering Research Express

Abstract

This paper presents the design and optimization of electrodynamic bearing (EDB) using Artificial neural networks (ANN). Initially, Finite element (FE) simulations were conducted to determine electric pole frequency ( ω rl ) , and the results of the simulations were fitted to an analytical model to obtain the stiffnesses (k1 and k2) and damping coefficients (c1 and c2) of an EDB. An ANN model was utilized for designing, optimizing, and predicting the stiffnesses, damping coefficients, and electric pole frequency ( ω rl ) of EDB. An ANN model with five different inputs, multiple types of algorithms, and different hidden neuron (HN) configurations was developed and trained to predict crucial bearing characteristics. The Bayesian Regularization (BR) algorithm with 10 HN demonstrated the lowest average error. Finally, the ANN model was used to optimize the EDB by utilizing MATLAB’s code of the Bonobo optimization algorithm. The optimization results were validated using the simulation results of COMSOL Multiphysics, and critical bearing characteristics were determined by curve-fitting an analytical model to the simulation outcomes. The results showed that ANN models could predict and enhance the EDB performance, as demonstrated by comparing results with the models’ predictions.

DOI

10.1088/2631-8695/ada7c5

Publication Date

3-31-2025

This document is currently not available here.

Share

COinS