A Novel Prognostic System for Predicting the Remaining Useful Life of Slurry Pumps that Exhibit High Fluctuation in Monitored Operating Signals
- Wai Tat Peter TSE (Principal Investigator / Project Coordinator)
- Jun Gong (Co-Investigator)
- Qing Zhang (Co-Investigator)
DescriptionSlurry pumps are commonly used in mining and heavy industry. A typical example is inthe production of synthetic crude oil extracted from oil sands containing around 60%sand, 20% clay, 15% bitumen and 5% water. The pump impellers are continuallybombarded by sand and small stones, and hence are prone to significant degradation dueto these adverse operating conditions. As the amount of sand and stones contained inthe pumped mixture is unpredictable, a temporal plot of the vibration magnitude showssignificant fluctuations. These fluctuations lead to difficulty in deriving slurry pumpdegradation trends, which are very different from the linear and obvious trendsassociated with pure fluid pumps. Hence, the prediction of the Remaining Useful Life(RUL) becomes difficult and inaccurate. To date, most of the reported methods of RULprediction are workable only for relatively linear trends. Few reports can be founddealing with the prediction of RUL generated from machines that exhibit a highfluctuation in their monitored sensor signals during operation.In 2007 there were 8,020 slurry pumps in China, and this number continues to increase.Slurry pumps are expensive to replace and generate the most severe fluctuations inmonitored signals. The frequent and unexpected failure of slurry pumps can interrupt oilproduction, cause serious contamination due to oil leakage from broken pumps or, in theworst case, cause human casualties. An unexpected shut-down of a slurry pump can cost$1.3 million per day in lost oil production. Hence, we aim to design a novel andintelligent prognostic system that can accurately predict the RUL of slurry pumpsoperating in an ever-changing operating environment. Such a prognostic system shouldbe self-adaptive and self-optimized in terms of prognostic parameters. It mustautomatically recognize relationships between the current operational status (health) ofthe monitored slurry pump and its simultaneously monitored sensor signals. Noprognostic system of this kind, based on operational data collected from on-site slurrypumps, has yet been designed.Once a successful system can be implemented for slurry pumps, the system can bemodified and applied to other types of machine that exhibit similar fluctuations insensor signals collected during operation. Potential candidates include reciprocatingmachines and car engines that also experience drastic changes in monitored signalsduring their operating. . The expected results are the minimization of productiondowntime, a dramatic decrease in maintenance costs and the avoidance of humancasualties.
|Effective start/end date
|1/01/16 → 12/12/19
- Fault diagnosis and prognosis,condition monitoring,remaining useful life,signal processing,