Application of rough set theory to fault diagnosis of check valves in reciprocating pumps

This paper investigates the application of rough set theory (RST) to fault diagnosis of check valves in three cylinder reciprocating pumps. Currently the faults are detected by the operator through readings of flowmeters and piezometers on the output pipeline. However the condition of each valve can not be detected in this way. Therefore a method of vibration measurements is used. A simulation pump is used to generate measurement data for several kinds of faults. To obtain the fault characteristics of the valves, the wavelet packet transform (WPT) is introduced as a preprocessing means of extracting time-frequency information from vibration signals. Furthermore, in order to determine the position of the faulty valve, RST is used in the process of fault diagnosis. It is proved that severe leakage and fault positions can be identified and diagnosed by the above method.