Research On Fault Diagnosis System Of Pumping Unit Based On Deep Learning

Rod pumping unit is an important equipment for crude oil exploitation,which is widely used in the process of oilfield development.When the pumping unit breaks down,it will not only affect the output and economic benefits of the oilfield,but also cause safety accidents in serious cases.Therefore,the fault diagnosis of the pumping unit is of great significance.At present,the mainstream fault diagnosis methods rely on engineers and technicians to identify the indicator diagram of pumping unit by virtue of experience.The workload is large and the efficiency is low,which can not meet the production needs of modern oil fields.Therefore,the research on advanced fault diagnosis algorithm,the development of intelligent pumping unit fault diagnosis platform and the real-time monitoring of pumping unit operation state have very important engineering application value and theoretical research significance.With the development of artificial intelligence technology,intelligent diagnosis technology based on deep learning is continuously applied to the field of pumping unit fault diagnosis.Generally,the traditional deep learning model has complex network structure and many parameters,and needs a lot of label data for training.However,the acquisition and labeling of pumping unit fault data will consume a lot of manpower and time,which restricts the training and optimization of the model.In addition,the generalization ability of the deep learning model is weak,and the difference between the training data and the actual data will also lead to the decline of the recognition accuracy of the model.Therefore,this paper uses generative countermeasure network to amplify the fault data of pumping unit,and improves the generalization ability of alexnet model through migration learning.The specific research contents are as follows:(1)Aiming at the problems of poor image quality and easy collapse of the model in the process of generating pumping unit fault data by conditional depth convolution generation countermeasure network(CDCGAN),a CDCGAN network model based on self attention mechanism is proposed.The model introduces self attention mechanism on the basis of CDCGAN,so that the network can pay more attention to the learning of important features in the process of feature extraction and reduce the influence of irrelevant information;At the same time,the regularization technology is used to consider the probability distribution of the generated image in the loss function to prevent the occurrence of mode collapse.The experimental results show that the CDCGAN model based on self attention mechanism can generate high-quality indicator diagram samples;By using the generated data to expand the small sample data set,the recognition accuracy of alexnet network for pumping unit faults is effectively improved.(2)Aiming at the problems of weak generalization ability of alexnet network model and poor effect of trained model in practical engineering application,a migration learning method based on pre training network is proposed for pumping unit fault diagnosis.Firstly,a large number of multi type source domain samples are used to train the alexnet network model to realize the preliminary optimization of network parameters;Then,fix the parameters of the feature extraction module of the alexnet network model,and train the classification module of the model with the pumping unit data.The experimental results show that the alexnet model after migration training can not only maintain the recognition ability of source domain samples,but also accurately identify pumping unit faults in practical application,enhance the adaptability to pumping unit fault samples,and improve the recognition ability of the model.(3)Based on the above research contents and combined with the actual needs of the oilfield,a pumping unit fault diagnosis system is developed.The system realizes the functions of data analysis,user management,real-time parameter display,fault diagnosis and historical data query.The system test shows that the system can diagnose pumping unit faults efficiently and accurately,give fault types and countermeasures,and the operation process is stable.