Tuan A. Z. Rahman, - IoT Systems Laboratory, MIMOS Berhad ۵۷۰۰۰ Kuala Lumpur, Malaysia
Leong Wen Chek - IoT Systems Laboratory, MIMOS Berhad ۵۷۰۰۰ Kuala Lumpur, Malaysia
Nordin Ramli - IoT Systems Laboratory, MIMOS Berhad ۵۷۰۰۰ Kuala Lumpur, Malaysia

The health state of rotating machinery can be represented by its vibration signal. With the help of machine learning approaches, the condition of rotating machinery such as electric motors can be diagnosed accurately. Thus, equipment down-time can be minimized and catastrophic accidents can be avoided. This paper presents an intelligent anomaly detection method for electric motors based on vibration signals. Due to lack of damage conditions information, only the normal condition data were employed to generate an unsupervised learning model for two different types of motor within the same class, which are the new laboratory motor and old industrial ones. The performance of the model to detect anomalies for both motors was studied, extensively. The results show that the model generated possesses the highest capability to detect anomalies for both motors as the normalized and mapped features were used.

anomaly detection, benchmarking, condition monitoring, machine learning, population-based