Purpose <p>Broken rotor bars (BRBs) are one of the industry's most common induction motor faults. However, choosing the best indicators to determine the measurement conditions for vibration analysis presents a challenge for the early location of this kind of fault.</p> Methods <p>This article addresses a monitoring and fault diagnosis approach based on the Taguchi method with grey relational analysis (GRA). This method is a commonly employed optimization technique for enhancing production quality. Nevertheless, we have adapted it to enhance the measurement conditions and determine the best measurement quality of vibration indicators. Three parameters, namely electric motor speeds, loads and type of sensor, are optimized to determine the most sensitive temporal features of vibration. A grey relational grade (GRG) obtained from the GRA is used to solve the early detection of rotor failure with multiple performance characteristics.</p> Results <p>Based on the GRG, optimum levels of measurement parameters have been identified, and their significant contribution is determined by the analysis of variance (ANOVA). Also, Linear and quadratic regression analyses were applied to predict the temporal features. The predicted values of the second-order mathematical model are very close to the measured ones and present a higher R square value. The performance of the quadratic regression model was validated by the high Pearson correlation coefficient and the low Mean Absolute Error.</p> Conclusion <p>The proposed approach has proven efficient in BRBs detection and it can be easily replicated for diverse motor defects.</p>