Faults Recognition System Of Pumping Unit Gearbox Based On BP Neural Network

As the most common transmission equipment in the oil production industry,pumping units directly affect the safety and efficiency of oilfield production.The gearbox belongs to the core component of the aerial part of the pumping unit.It is a complex equipment that integrates electrical,mechanical and transmission professionals.At the same time,it is also a lossy device with a high failure rate,so the pumping unit gearbox is selected as the research object.The faults identification and diagnosis research on the gearbox and rolling bearing of gearbox is carried out to realize the accurate identification of the fault by taking the fourth oil production plant production preparation brigade pumping unit overhaul workshop of Daqing Oilfield as test platform.In this paper,by collecting the fault vibration information of the gearbox,the key technologies such as signal detection,denoising preprocess,feature extraction,resonance demodulation and fault type identification involved in the fault diagnosis of the gearbox are mainly studied.Through a large number of test verification,Based on the wavelet packet algorithm and the BP neural network intelligent algorithm,the fault type is accurately identified.The LabVIEW technology is used to construct the gearbox fault diagnosis system of the pumping unit.The system has good practicability and can meet the maintenance requirements in the production process.This paper takes a lot of related research work on the basis of fault mechanism and signal processing.The main contents are as follows:(1)Firstly,the internal structure of the pumping unit gearbox is analyzed,on the basis of which,the typical fault categories and manifestations of gears and rolling bearings are deeply analyzed.At the same time,the vibration mechanism of gears and rolling bearings and the corresponding fault characteristics information are clarified,and based on this theory,the research ideas for solving the problem of gearbox fault diagnosis are established.(2)Aiming at the problem of fault feature extraction of pumping unit gearbox,several analysis indicators involved in time domain statistics are discussed in depth.By testing multiple times on several testing points,the preferred test point is selected and the root mean square and peak factor are selected at the same time.The parameters such as the margin factor and the kurtosis factor which can better reflect the fault characteristics are used as the characteristic parameters of the time domain analysis of the gear and the rolling bearing,and the fault category is initially confirmed.In the frequency domain,the Fourier transform is used to obtain the natural frequency of the fault signal and the frequency band range of the side frequency.In order to eliminate the interference from noise,the resonance demodulation feature extraction technique based on wavelet packet algorithm is studied.The specific process is using the wavelet packet algorithm to decompose the fault information firstly,and then select the low-frequency wavelet packet coefficients for signal reconstruction,and finallytake resonance demodulation analysis on the reconstructed signal to find the response characteristic frequency of the fault.(3)Based on the basic principle of BP neural network algorithm,the gearbox faults recognition model is constructed.Meanwhile,the basic parameters such as the input layer node and the output layer node of the neural network are set,and the gear working conditions and different working conditions of the rolling bearing are respectively coded to establish clear rules for the determination of the type of faults.The neural network fault identification system is sampled and trained by the working condition data of gears and rolling bearings provided by the oilfield maintenance brigade test platform to realize the recognition of the normal state,broken teeth and abrasion state of gear.As for rolling bearings,normal state,states recognition of outer ring faults,inner ring faults and ball faults can be achieved.The recognition accuracy of the system is as high as 95%,which can effectively identify some types of faults occurring in oil field production process,and has good practical application value.(4)Based on the theoretical analysis,LabVIEW is used to develop the platform of the gearbox fault diagnosis system,and the fault diagnosis interface is designed.And combined with Matlab,the time-frequency domain analysis,the resonance demodulation processing based on wavelet packet and faults identification based on BP neural network are realized.The fault diagnosis results are displayed on the interface which is simple and clear,and the good human-computer interaction characteristics of the fault identification system are taken into account.What's more,the system also has good portability and functional expandability.