Hybrid Approach Based On Tensor Decomposition For Model Order Reduction Applied To Motor Diagnosis

ABSTRACT Accurate and efficient diagnosis of motor eccentricity faults is crucial for the reliable operation of electric machines. This requires large volumes of high‐quality data across various operating conditions and fault types. While experimental data collection is costly and time‐consuming, simulations offer a more practical alternative. However, full‐order models are often computationally prohibitive due to magnetic saturation nonlinearities and repeated simulations across different conditions. This challenge motivates the development of efficient model order reduction (MOR) techniques. To alleviate the computational cost associated with nonlinear problems, look‐up table (LUT) interpolation can be used to approximate nonlinear operators, thus avoiding convergence issues commonly encountered in hyper‐reduction methods. However, conventional LUT methods face significant memory demands, especially in multi‐parameter settings. To address these limitations, a hybrid MOR framework is proposed, based on a two‐stage approach leveraging tensor decomposition, tailored for nonlinear, multi‐parametric electromagnetic problems in motor diagnostics. The two‐stage strategy is applied to both the LUT interpolation part and the MOR process. The framework is validated on a permanent magnet synchronous motor with static eccentricity, demonstrating superior accuracy and efficiency compared to traditional MOR techniques, and showing great potential for surrogate modeling in motor fault diagnosis.