Practical aspects of rotor cage fault detection for medium-voltage induction motors

Predictive diagnostics offering early failure detection of large induction motors applied in metals, pulp & paper and other process industries are becoming increasingly important. As motors grow larger, industry has become increasingly reliant on technologies to detect rotor faults via on-line prognostics and arrange optimal maintenance intervals to increase productivity. Traditional broken rotor bar fault detection algorithms have historically relied largely on monitoring changes in the stator current spectra. This often results in nuisance warnings when the motor operates at different load levels, or when baseline data at healthy motor operations are not available. To address this issue, a fault severity evaluation technique is introduced in this paper to detect rotor cage failures using only current and voltage measurements, plus selected motor nameplate data and motor´s geometric dimensions. The fault severity index can indicate the possibility of a rotor cage fault even in the absence of baseline data. This guarantees the algorithm´s reliability in practical applications. In addition, a decision-making system, including an adaptive filter and fuzzy logic, is proposed to warn the user in the case of a rotor cage failure. Experimental results show that the proposed fault severity evaluation algorithm can reliably reflect the rotor cage status under different operating conditions, which can be further applied in the detection of rotor cage failures.