Track: Data Analytics and Big Data

Abstract

Failure Detection and diagnosis is critical for maintaining the health of a system. Therefore, most of the production industries today try to follow Condition Based Maintenance (CBM) which works on the strategy that monitors the actual condition of the asset to decide what maintenance needs to be done. CBM monitors the health of the mechanical components through measurements of different kinds of sensor data such as vibration, temperature, and pressure, etc. By using this status data, it can be determined whether maintenance is required or not indicated by decreasing performance or upcoming failure. Compared with preventive maintenance, this increases the time between maintenance repairs, because maintenance is done on an as-needed basis. In this research, we employ Hidden Markov Models (HMM) to recognize the different health states of a mechanical component. The transition between states is probabilistic in HMM which helps in estimating the future state of the component. The proposed health state estimation method has been validated on a time-series sensor signal of compressors.