Fault classification improvement in industrial condition monitoring via Hidden Markov Models and Naïve Bayesian modeling

Polyphase induction motors are the most commonly available industrial machines utilized in a wide range of real-world applications. Any impending fault within these motors is generally very difficult to isolate by conventional fault sensors or experts. The fact is generally attributed to the non-linear behavior of the motor's terminal characteristics. Extracting anomalous behavior from such data is a challenging task and predominantly relies on the historical machine data pattern. Based on the abovementioned context, this paper presents a novel time-series condition monitoring data assessment methodology to identify developing faults in induction motors. The technique employs Hidden Markov Models to identify anomalous machine behavior. The identification thus obtained is further improved via a Naïve Bayes classifier to further eliminate false positives from healthy and fault-containing data. The overall Bayes classification outcome showed a marked increase in detection accuracy at 84.55% with a substantial reduction in false positives.