This project assessed the behaviour of a positive displacement type of compressor utilising airborne acoustic signatures. The study concentrated on finding an improved method based on airborne sound that can be suitable for diagnosing some common faults in reciprocating compressor (RC). Being a critical component of the industry, the condition monitoring of reciprocating compressor is very much needed to avoid any failure of its machine parts that can cause a sudden breakdown of RC. The compressor acoustic signal is a result of various mechanical forces related to the varied cylinder pressure, valve movement, turbulence air flow which in terms contribute to the periodic excitation along with the non-linearity caused by the valve fluttering, hence making the airborne signal complex and non-stationary in nature. The transient response due to the periodic impact of the valves, modulation effect due to the fluid-mechanical interaction and low signal to noise ratio (SNR) are the challenging aspects of this study. To demonstrate the vibro-acoustic property of the reciprocating compressor, first a model was developed. The leakages in valve and intercooler are very common in RC. The second most common fault which is often neglected is a clogged filter. Hence, taking into consideration, filter blockage fault is introduced for the first time in the existing test set up. Three faults (discharge valve leakage, intercooler leakage and filter blockage) are simulated, and corresponding acoustic responses are recorded for further study of the signal-nature. The model is then validated by the actual data from RC test bed. Along with the modelling of compressor acoustics, various signal processing techniques like Minimum Entropy Deconvolution (MED), Teager Energy Operator (TEO) are used on the test data to detect abnormalities present. MED in this case, is proved to be effective in finding the transient responses whereas, TEO serves as an energy detection tool for tracking the total mechanical energy. Still both methods find it difficult to come up with the best possible diagnosis results as they fail to take all the major characteristics of the RC acoustics into consideration. To overcome this challenge, higher order spectral analysis as a form of Modulation Signal Bi-Spectrum (MSB) is used to find out the most effective modulating components by enhancing the modulating characteristics and suppressing the noise. Moreover, the quadratic phase coupling allows MSB to handle the non-linearity that might be present in RC due to the valve fluttering. The proposed MSB based method not only provides a more consistent and accurate diagnosis of compressor faults but also shows that airborne acoustics has a good aspect in fault identification of RC by validating both model and test results. Recognizing that there is perpetual room for improvement, the performance of the proposed RC fault diagnosis method can be enhanced by incorporating a denoising technique developed using the Variational Mode Decomposition (VMD) associated with Kalman filtering method. The future study must also consider several other individual and compound faults that can be incorporated in the study for understanding vibro-acoustic phenomena of RC.
| Date of Award | 23 Apr 2024 |
|---|
| Original language | English |
|---|
| Supervisor | Andrew Ball (Main Supervisor) & Fengshou Gu (Co-Supervisor) |
|---|