A robust method to monitor the operating conditions of induction motors is presented. This method utilizes the data analysis of the air-gap torque profile in conjunction with a Bayesian classifier to determine the operating condition of an induction motor as either healthy or faulty. This method is trained offline with datasets generated either from an induction motor modeled by a time-stepping finite-element (TSFE) method or experimental data. This method can effectively monitor the operating conditions of induction motors that are different in frame/class, ratings, or design from the motor used in the training stage. Such differences can include the level of load torque and operating frequency. This is due to a novel air-gap torque normalization method introduced here, which leads to a motor fault classification process independent of these parameters and with no need for prior information about the motor being monitored. The experimental results given in this paper validate the robustness and efficacy of this method. Additionally, this method relies exclusively on data analysis of motor terminal NOT THE PUBLISHED VERSION; this is the author's final, peer-reviewed manuscript. The published version may be accessed by following the link in the citation at the bottom of the page.