Dynamic mode decomposition based fault diagnosis in three-phase electrical machines

Three-phase electrical machines are widely used in various industrial applications, and the mechanical fault in those machines leads to an oscillation in the load torque, which introduces an amplitude and/or phase modulation in the stator current. This work proposes a methodology to detect mechanical faults in rotating machines using dynamic mode decomposition (DMD). The DMD technique utilises singular value decomposition (SVD) to extract the dominant modulation frequencies and indexes by constructing a shifted-stack matrix of the three-phase current signal. Further, to validate the proposed methodology, an unbalance in the three-phase signal is examined, which includes both amplitude and phase unbalance with additive white Gaussian noise. The results demonstrated that the proposed DMD extracts the modulation frequencies and indexes with a maximum error of 2%. Additionally, different fault severities are considered to establish a real-time scenario, and the results show that the proposed DMD effectively identifies the faults with a maximum error of 1.3%. • DMD is applied to estimate the modulation index along with the modulation frequency. • An independent investigation assesses the performance of DMD with amplitude and phase-modulated signals. • A mixture of modulated and non-modulated components in the stator current is considered to verify the efficacy of the DMD. • Different severity of the faults such as S 1 , S 2 , S 3 , and S 4 are considered with additive noise.