A novel implementation of vibration signal decomposition for estimation of degradation in rotating plant

Effective and transparent monitoring of rotating plant assets is essential to the continued reliable operation of power stations. Rotating plant monitoring generally includes analysis of vibration signals, where operations and maintenance engineers use the output from vibration sensors to justify the continued operation of the plant or plan for maintenance interventions where necessary. One common approach to such vibration monitoring is the adoption of alarm driven strategies where certain operational or mechanical interventions are performed when thresholds are triggered due to deviations from a predefined operational envelope. This reactive intervention approach, however, does not provide operators or equipment manufacturers with any insight into the long-term degradation of a rotating plant item, which could be used to mitigate unplanned stoppages. This paper proposes the novel implementation of Empirical Mode Decomposition to boiler feed pump vibration signals, alongside subsequent statistical analysis of the decomposed signals to estimate time-frames associated with alarm violations and entry into predefined zones of operation. Such a technique provides pump operators with information that can be used to plan for future maintenance interventions and pump manufactures with insight into the likely degradation of their product during sustained operation.