Non-intrusive electrical and mechanical fault diagnosis of induction motors via Fourier-transform-aided learning

Abstract

Induction motors, which are widely used in domestic and industrial applications, often operate under harsh mechanical, electrical, and thermal conditions that can lead to unexpected faults. Effective condition monitoring and fault diagnosis are therefore essential to ensure operational safety, reliability, and productivity. While recent advances in artificial intelligence have accelerated the adoption of data-driven monitoring solutions, the limited explainability and poor generalization of black-box models remain major concerns for industrial deployment. To address this issue, this paper proposes a knowledge-based input configuration to inform deep learning models for both electrical and mechanical fault diagnosis, rather than increasing model complexity. Using electrical measurements and the fast Fourier transform (FFT), only 11 fault-related frequency components are extracted based on motor parameters, instead of computing the full FFT spectrum with thousands of components. This drastic reduction in feature size significantly lowers computational and memory requirements, making the method more suitable for deployment on resource-constrained platforms. The proposed approach is evaluated against three benchmarks: raw electrical signals, wavelet transform–based time–frequency features, and principal component analysis–based FFT dimension reduction. Results confirm that, while conventional feature processing techniques perform well for electrical fault diagnosis, only the proposed FFT-informed input effectively captures both electrical and mechanical fault patterns and maintains strong generalization across independent experiments. By integrating the selected FFT components and a cascading workflow, the framework achieves an overall testing accuracy of 98.28%, in contrast to 59.01% when FFT features are excluded and the diagnosis system operates as a simple 4-class classifier.