Compound Fault Diagnosis of Centrifugal Pumps Using Vibration Analysis Techniques

Centrifugal pumps are widely used in many different industrial processes, such as power generation stations, chemical processing plants, and petroleum industries. The problem of failures in centrifugal pumps is a large concern due to its significant influence on such critical industries. Particularly, as the core, parts of a pump, bearings and the impellers are subject to different corrosions and their faults can cause major degradation of pump performances and lead to the breakdown of production. Therefore, an early detection of these types of faults would provide information to take timely preventive actions.

This research investigates more effective techniques for diagnosing common faults of impellers and bearings with advanced signal analysis of surface vibration. As overall vibration responses contain a high level of broadband noises due to fluid cavities and turbulences, noise reduction is critical to developing reliable and accurate features. However, considering the modulation effect between the rotating shaft, vane passing components and any structural resonances, a modulation signal bispectrum (MSB) method is mainly used to extract these deterministic characteristics of modulations, which differs from previous researches in that the broadband vibration is often characterised with statistical methods, high frequency demodulation along spectrum analysis. Both theoretical analysis and experimental evaluation show that the diagnostic features developed by MSB allow impellers with inlet vane damages and exit vane faults to be identified under different operating conditions.

It starts with an in-depth examination of the vibration excitation mechanisms associated with each type of common pump faults including impeller leakages, impeller blockages, bearing inner race defects and bearing outrace defects. Subsequently, fault diagnosis was carried out using popular spectrum and envelope analysis, and more advanced kurtogram and MSB analysis. These methods all can successfully provide correct detection and diagnosis of the faults, which are induced manually to the test pump.

Envelope analysis in a bands optimised with Kurtogram produces outstanding detection results for bearing faults but not the impeller faults in a frequency range as high as several thousand hertz (about 7.5kHz). In addition, it cannot provide satisfactory diagnostic results in separating the faults across different flow rates, especially when the compound faults were evaluated. This deficiency is because they do not have the capability of suppressing the random noises.

Meanwhile, it has found that the MSB analysis allows both impeller and bearing faults to be detected and diagnosed. Especially, when the pump operated with compound faults both the fault types and severity can be attained by the analysis with acceptable accuracy for different flow rates. This high performance of diagnosis is due to that MSB has the unique capability of noise reduction and nonlinearity demodulation. Moreover, MSB diagnosis can be a frequency range lower than 2 times of the blade pass frequency (<1kHz), meaning that it can be more cost-effective as it demands lower performance measurement systems.

In addition, the study also found that one accelerometer mounted on the pump housing is sufficient to monitor the faults on both the impeller and the bearing as it uses a lower frequency vibration which propagates far away from the bearing to the housing, rather than another accelerometer on the bearing pedestal directly.