Fault detection and identification for centrifugal compressor by ensemble model

In this work, a machine's health monitoring system (MHMS) that can detect anomalies for a centrifugal compressor, operating in a real plant, is proposed. A compressor is a crucial machine but over time its performance can decline, and faults can develop. In the compressor under investigation, several sensors have been installed. Subsequently, its behavior is monitored and recorded. During this period, no noticeable fault is detected. In the proposed technique, simulated faults are employed to build up the data that can be used in the investigation, then the performance of the compressor is analyzed and compared with ML and MLP. Herein, the study indicates that the data accumulated is a good candidate for this challenging case. Overall, the proposed technique demonstrates good potential for detection of anomalies regarding the real centrifugal compressor.