Fault Feature Extraction And Diagnosis Of Reciprocating Pump Based On HHT

In the recent years, as the proportion of mechanic industry in nation economy increases, the requirement for safe operation of mechanic equipment is higher and higher.Due to that all kinds of technology are progressively applied to mechanic equipment fault and diagnosis,such as signal processing,pattern recognition,etc.,mechanical equipment fault and diagnosis technology has been developed rapidly all over the world.The reciprocating pump as a part of mechanic equipment field takes up very important station in production department,such as the petrochemical industry.Because its applications are very widely, making researchs about its fault and diagnosis has important meaning.Though people have already made out some researchs and obtained some research results, but the total diagnosis level is not still very high.Based on former research results, according to the strong non-stationary characteristic of the pump valve signal of the reciprocating pump, this text introduces Hilbert-Huang Transform (abbreviated as HHT), and diagnosed the pump valve faults of the reciprocating pump well.Because reciprocating pump has complicated structure and more exciters, so its signal is a strong non-stationary signal, and carrying out fault feature extraction and diagnosis is very difficult to it, this text mainly researches on featute extraction of reciprocating pump's valve vibration signal.This text introduces HHT that Huang put forward , it is a kind of signal processing method that suits for dealing with the stationary signal, and suits for non-stationary signal also.Although the Hilbert-Huang Transform (HHT) is an effective tool processing the non-stationary signal, the HHT based on Emperical Mode Decomposition (abbreviated as EMD) algorithm which adopts the cubic spline interpolation couldn't acquire accurate characteristics for the strong non-stationary signals in that the spline produces an accurate result only under the condition that the data consists of values of a smooth function. To solve this problem, the paper presents the Hilbert-Huang Transform (HHT) based on the new improved EMD algorithm which adopts Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) as envelop of extrema. Through the simulation comparison between the PCHIP and the spline interpolation, it can be shown that the PCHIP has some advantages, such as no overshoots and less oscillation, which the spline interpolation doesn't have when a signal is not smooth. On the other hand, the PCHIP can more reflect the trends of the data. Applying these two methods to the same signal measured on reciprocating pumps valves separately, it can be found that the EMD based on the PCHIP can more accurately decompose the highly non-stationary signal than the spline interpolation, which can afford a good measure to extract the fault characteristics of the signal. As described above, the method proposed in this paper can effectively diagnose the fault of the valves with spring failure.