The characterization of incipient failure patterns for three-phase induction motors and the development of classification methods for predictive maintenance applications in the mining industry

A methodology for the identification and classification of patterns of incipient failure due to insulation degradation in the stator windings of induction motors is presented. A mathematical model of an induction motor is used to build a database of patterns for different types and stages of deterioration. Deterministic and statistical methods of pattern recognition are applied to develop viable pattern classification schemes for both wye- and delta-connected stator windings.;Patterns for the identification of phase-to-phase faults in wye-connected motors have been found, and automated recognition for this type of faults appears possible using linear discriminant functions. Recognition of deterioration patterns for phase-to-phase faults in delta-connected motors may be achievable using principal component analysis. A regression model which establishes a relationship between negative-sequence current and fault power is proposed as mean to assess fault severity in wye- and delta-connected motors.