Airborne Acoustic Signature Analysis for Fault Diagnosis of Reciprocating Compressors Using Modulation Signal Bi-spectrum

Debanjan Mondal, Usama Haba, Fengshou Gu, Andrew Ball

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

9 Citations (Scopus)


Reciprocating compressors are the most important part of a petrochemical industry. In order to monitor the conditions of this complex machine, a remote fault diagnosis technique based on airborne acoustic signature analysis has been proposed. However, as there are many rotating and reciprocating components involved, extracting the characteristic features from the non-stationary and non-linear acoustic signals resulted by those machine components are very difficult. The presence of structural or acoustic resonance may further contribute to the signal modulation which makes the acoustic signal of the compressor very complex. In this paper, the modulation signal bi-spectrum method has been applied to the compressor sound signals with the capabilities of suppressing random noise, demodulating non-modulation components, and estimating modulation degrees. It allows an in depth representation of the non-linear effects of the modulation signals due to the repetitive valve impacts, airflow fluctuations, resonance phenomenon, and thereby providing a more accurate diagnosing feature to identify the root cause of the faults. The experimental study examines various kind of reciprocating compressor (RC) faults including intercooler leakage, discharge valve leakage and filter blockage. The analysis results show the effectiveness of the proposed method in diagnosis of these faults based on the airborne acoustic signature analysis.


Conference25th IEEE International Conference on Automation and Computing
Abbreviated titleICAC 2019
Country/TerritoryUnited Kingdom
Internet address


Dive into the research topics of 'Airborne Acoustic Signature Analysis for Fault Diagnosis of Reciprocating Compressors Using Modulation Signal Bi-spectrum'. Together they form a unique fingerprint.

Cite this