Research On Fault Diagnosis System Of Pumping Gearbox

Pumping is one of the most commonly used transmission equipment in the oil industry. It’s a complex machinery that concentrates machinery, electrical, mechanical transmission, hydraulic, electronics, etc.The normal working condition of pumping system directly affects the safety and production of oil exploration.So the study of condition monitoring and fault diagnosis technology of oil pumping have significance and theory value to prevent pumping system from sudden accidents.In this paper, take pumping gearbox as the research object. In view of the key problem of the gearbox vibration signal denoising and feature extraction, study the method suitable for pumping gear vibration signal de-noising, processing and identification. And using the method of energy distribution based on relatively generalized S transform to dispose the problem of gearbox signaling. Use the measured data of pattern recognition to identify all kinds of gear box fault effectively, and examine the feasibility of the scheme. According to the actual production of the oil industry, design and develop the pumping gearbox fault diagnosis expert system of distance changing kernel function combining with semi-supervised local embedding manifold learning algorithms. This paper conducts a series of studies through the preprocessing of vibration signal.And The main contents are as follows:Firstly, it is necessary to analyze the vibration mechanism of pumping gear and to summarize the research status of rolling bearing diagnosis information acquisition, fault feature extraction, pattern recognition and other facts research status at home and abroad. On this basis, establishes the general ideas and research content of this article.And then, in view of the big disturbance of pumping gearbox vibration signal acquisition, the paper presents a hybrid echo noise cancellation technology composed by SVM and hyperbolic tangent function LMS algorithm which adds geometric decline factor. Through theoretical analysis and simulation comparison, the results show that the method can abtain more obvious signal characteristics. And the improved algorithm has many advantages such as smaller steady-state error, the ideal convergence speed and higher signal-to-noise ratio. The algorithm applied to the actual vibration signal processing of pumping gearbox, fully meets the requests of gearbox vibration signal noise cancellation. Providing a reliable guarantee for further signal feature extraction.Based on the analysis of pumping gearbox fault signal, the research of traditional methods: Fourier, Hilbert transform and wavelet transform is conducted. Through the study of theoretical and simulations, find the characteristics and deficiencies of these traditional algorithms. Based on the generalized S transform,this paper puts forward the relatived generalized S transform energy distribution method to extract vibration signal characteristics of pumping unit gearbox. Researching the theoretical basis and the implementation process of this method, analysis broken teeth, cracks, sharp teeth failure of gearbox by simulations of synthetic noise signal. This method is intuitive and effective, giving a good solution for the problem of not fully reflecting the characteristic information at characteristic frequencies. and providing a reliable source of data for further intelligent pattern recognition.According to the characteristics of the traditional semi-supervised learning algorithms, this article, through the distance change and combine embedded kernel function, improves the semi-supervised learning algorithm locally linear embedding. The simulation of synthetic signal and measured signal verify the validity of the improved algorithm. Using the improved semi-supervised learning algorithm locally linear embedding to diagnose and analysize the vector combined by characteristics of time domain and frequency domain energy distribution information features can reflect the intelligence and effectiveness of the intelligent fault diagnosis platform.Finally,Based on the improved echo de-noising and relatively generalized S transform energy distribution and energy entropy feature extraction method, the paper designs a pumping gearbox fault diagnosis system with LabVIEW. The fault diagnosis system consists of five basic modules, signal playback module, time-domain analysis module, frequency domain analysis module, the frequency analysis module based on relatively generalized S transform energy features and fault diagnosis system module based on improved semi-supervised learning algorithm locally linear embedding. The measured diagnostic results show that the design of the scheme is feasible and have great significance for the safety of production.