DiagramDataLiBowen.

At present, the fault diagnosis of pumping units in major oil fields in China is time-consuming and inefficient, and there is no universal problem for high requirements of hardware resources. In this study, a fault fusion diagnosis method of pumping unit based on improved Fourier descriptor (IDF) and rapid density clustering RBF (RDC-RBF) neural network is proposed. Firstly, the minimum inertia axis of the center of gravity of the indicator diagram is obtained. The farthest point of the intersection of the inertial axis and the contour is determined as the starting point. Then Fourier transform is performed on the contour boundary of the graph to obtain the feature vector. Then, combining with the idea of fast density clustering algorithm, the number of hidden layer neurons of RBF is determined by finding the point with the highest density and using it as the hidden layer neuron. At the same time, the characteristics of Gaussian function are introduced to ensure the activity of hidden layer neurons. Finally, through dynamic adaptive cuckoo search (DACS), the step size is automatically adjusted according to the convergence speed of the objective function of RBF, and the efficiency and accuracy of RBF in different search stages are balanced. The optimal parameters such as the width and weight of RBF are determined, and the optimal RDC-RBF fault diagnosis model is established. The model is applied to the diagnosis of different fault types of pumping units, and compared with the current mainstream models. The average detection accuracy of the fusion RDC-RBF fault diagnosis method proposed in this paper reaches 96.3%. The measured results have high accuracy and short time. At the same time, this method is currently applied to oil production sites such as Shengli Oilfield in China, which greatly reduces the human resources required for fault diagnosis of pumping units in the past.