LIU Jinhai, SHI Liming, KONG Ganlin, et al. Integrated Open-circuit Fault Diagnosis Strategy for Multi-phase Motors Based on Extended Dice Similarity Coefficient[J]. 2026, 46(11): 4776-4786.
DOI:
LIU Jinhai, SHI Liming, KONG Ganlin, et al. Integrated Open-circuit Fault Diagnosis Strategy for Multi-phase Motors Based on Extended Dice Similarity Coefficient[J]. 2026, 46(11): 4776-4786.DOI: 10.13334/j.0258-8013.pcsee.242483.
Integrated Open-circuit Fault Diagnosis Strategy for Multi-phase Motors Based on Extended Dice Similarity Coefficient
This paper proposes an integrated open-circuit fault diagnosis method for multi-phase motors
enabling rapid fault detection under high-dynamic operating conditions such as acceleration/deceleration and time-varying root mean square (RMS) phase current. First
a fault indicator is developed based on the motor current components (ix
iy) in the harmonic plane to identify the occurrence of open-circuit faults in the system. Subsequently
according to the relationships between phase currents
fundamental currents
and harmonic currents
Type-I fault indices for each phase of the motor are designed using the extended Dice similarity coefficient. Based on the connection topology of the motor stator windings
the constraint relationships of phase currents are derived
and Type-II fault indices for each phase are further formulated. By integrating the characteristics of the two types of fault indices
similarity coefficient fault indices Dj for each phase are constructed to online locate the specific faulty phase. The proposed fault indicator and indices are normalized and independent of the fundamental frequency of phase currents
which makes them robust against sudden changes in RMS phase current and time-varying speed
thereby achieving independent fault diagnosis for each phase. Finally
simulations and experiments are conducted on a dual three-phase permanent magnet linear synchronous motor (PMLSM) test platform to validate the effectiveness of the proposed method. Experimental results demonstrate that the proposed strategy can achieve fault detection within 7.89 ms (half a fundamental cycle)
reducing the diagnosis time by 50% compared with conventional methods
and the diagnosis results are not affected by variable speed