Diagnosing of air compressor faults using frequency data driven approach

Faults in rotating or reciprocating machines can be induced due to lack of maintenance or pre-existed defect of the components during the manufacturing. These faults can cause dangerous consequences and great cost losing or even cause fatal incidents. Thus, detection of these faults in a suitable time will no doubt provide the opportunity for the decision-makers to determine the suitable intervention and to avoid the possible break down. In the present research, a frequency data-driven based methodology is developed to conduct the detection and identification of faults presence and type in an air compressor. This methodology has two main stages; feature extraction and classification. The features are extracted from the frequency representation of signals and each feature represents the mean value of a certain frequency band amplitude. A deterministic similarity measure is used in the classification stage. The results show that the present simple methodology can effectively detect and identify the air compressor faults.