Decision Support System for Diagnosing Broken Rotor Bar Severity in Induction Motors

Unexpected failures in induction motors can cause serious disturbances in production lines, leading to significant financial losses. Fortunately, incipient faults can often be predicted before they evolve into critical failures, allowing for preventive maintenance. This paper presents an intelligent decision support system for diagnosing broken bar severity in three-phase induction motors, identifying the number of broken bars. A well-established technique for fault diagnosis is Motor Current Signal Analysis, which identifies faults in the frequency domain. The proposed system advances the state of the art by automatically identifying the best sideband that indicates the defect and choosing the minimum required signal acquisition time. Whereas most recent studies pre-select the sidebands of the fundamental, $5^{\mathrm {th}}$ , or $7^{\mathrm {th}}$ harmonics, this study analyses and uses the best sideband of the fundamental, $3^{\mathrm {rd}}$ , $5^{\mathrm {th}}$ , $7^{\mathrm {th}}$ , $9^{\mathrm {th}}$ , and $11^{\mathrm {th}}$ harmonics, considering the first three sidebands, to improve detection accuracy. Moreover, the decision support system makes this decision by taking into account the motor load condition and sets the shortest acquisition time that ensures a significance level of less than 5%. Results show that the first left sideband of the $3^{\mathrm {rd}}$ harmonic is often the most effective indicator, except during low load conditions, when other sidebands show superior accuracy. In addition, as the load increases, the time required for reliable frequency-based fault detection is reduced to 15 periods, or 0.25 s, suitable for loads above 37.5%. This capability enables near real-time applications and dynamic predictive maintenance strategies, aligning with Industry 4.0 and industrial digitalization paradigms.