Research On Pumping Fault Diagnosis Based On Artificial Fish Swarm Algorithm And Neural Network

In the oil exploration and production,a large number of oil pumping machines and other production equipment are field operations. They are geographically dispersed, which surrounding environment is usually bad and difficult to have manual inspection. The pumping oil under complex environment often occurs faults. It is not conducive to underground mining equipment and process real-time monitoring, thus affecting oil production and efficiency. Therefore, timely and accurate diagnosis the fault of pumping unit is helpful to improve the oil recovery efficiency and reduce production cost.First, based on the analysis of the indicator diagram of the original features extraction method, the indicator diagram theory are introduced in this passage. And for various fault types of pumping unit, we select some parts to introduce. Then, the pretreatment of the indicator diagram and image segmentation in the optimal threshold for image boundary to get maximum filling area. Proposed a method combining the moment invariants and Fourier descriptors of pumping units is the fault diagnosis indicator diagram. After computing moment invariants of images, we can obtain the moment invariants sequence, which can instead of the images. Through the discrete Fourier transform to obtain normalized moments Fourier descriptors with translation, rotation and scaling invariance. By Euclidean distance method to verify the speediness and effectiveness.Secondly, discussing the AFSA and artificial neural network algorithm. BP neural network is easy to fall into local extreme. And the convergence is slow. For the fish based on improved BP neural network to fault diagnosis of pumping dynamometer. Set up artificial fish BP neural network model using the Matlab simulation software. This paper put forward a proposal to verify the speediness and effectiveness of the comparative analysis of simulation results.The research not only enriches the method of pumping dynamometer feature extraction, and to realize the fault diagnosis of oil pumping machine theory broadens the scope.