Study On Hydraulic Terminal Fault Diagnosis Of Reciprocating Pump

Reciprocating pump as the important reciprocating equipment in the industrial fields is widely used in oil drilling, oil and gas exploration production areas. In the use of the fluid end which is the weak link of reciprocating pump failed frequently, due to the complex structure of fluid end, the excitation source, leading to the fault diagnosis of hydraulic end is more difficult. So, the fault diagnosis of reciprocating pump fluid end has a very important significance for improving the efficiency in the use of reciprocating pump and to ensure the safe production. Plunger wears failure and valve component failure is key points for diagnostic studies in this paper. By using the time-domain statistical indicators diagnostic method, piston and cylinder wear fault are diagnosed. Meanwhile, the fault characteristics of the valve components was extracted by use the wavelet packet analysis method, BP neural network was created in MATLAB, and the intelligent recognition of valve component failure was achieved by the neural network.These diagnostic schemes are good for fault diagnosis of reciprocating pump fluid end.This research work is mainly reflected in the following aspects:First, based on the extensive research of foreign literature and field data, the work principle of fluid end of reciprocating pump, common faults and cause of the malfunction are analyzed in this paper. The streamlined bonnet vibration mechanics model was established.The feasibility of using vibration signals for fault diagnosis of hydraulic end was discussed. The failure and non-stationary of fluid end vibration signal are major causes of the changes of vibration signal. The experimental program of the failure of hydraulic side is designed. There are failure experiments in3DS-1/12.5reciprocating pump.Second, aimed at the characteristics of non-stationary vibration and the transient signal for the fluid end, the time-domain statistical indicators diagnostic method was used for the plunger wear fault. And the Important coefficient of the various indicators was calculated. The piston cylinder liner degree of wear of the diagnostic models was established. The value of the fault degree is more intuitive reflect the degree of the plunger wear. Aimed at the valve vibration signal characteristics of non-stationary and difficulty of fault feature extraction, the wavelet and wavelet packet transform method was proposed to make wavelet packet analysis of vibration signals based on comparative analysis of several time-frequency analysis methods. Maximum energy difference method was proposed to determine the wavelet packet decomposition program. The different types and extent of failure characteristics of the valve component were effectively extracted by the establishment of "frequency-energy-failure " diagnostic model.Then, the principle of BP neural network algorithm is detail discussed in the paper. The creation and training of the BP network are achieved by the improvements measure Of BP network in MATLAB. The intelligent recognition of the valve component failure was achieved by the created neural network, and the accuracy rate of diagnostic was more than 90%.Finally, for the existing problems of experiment, faults diagnosis program of reciprocating pump fluid end and the overall performance diagnostic programs were designed. And the testing hardware system program of the fluid end of reciprocating pump was designed. The implemented of diagnostic test method has improved in the oil field.