Application of CNN-SVM model in fault diagnosis of pumping well
ZHONG Zhidan
FAN Haojie
LI Penghui
Abstract：In the fault diagnosis of pumping well by the traditional identification methods of indicator diagram, there are two main problems,manual selection of indicator diagram features and low identification accuracy. Based on artificial Intelligence theory,an intelligent identification model of indicator diagram is proposed by the combination of convolution neural network ( CNN) and support vector machine ( SVM).The features of indica-tor diagram are automatically extracted by means of convolutional neural network,and according to the extrac-ted deep image features,the results of fault diagnosis can be obtained from support vector machine.The experi-mental results show that the combination of CNN and SVM not only omits manual selection of indicator diagram features,but also increases the identification accuracy to 99.71%.The test performance is superior to other i-dentification models.A feasible solution is provided by the proposed model for the rapid and accurate fault di-agnosis of pumping well,which is significant for the efficient operation of oil fields.
Keywords：convolutional neural networksupport vector machineindicator diagram identificationfault diag-nosisdeep learning
Publication Date：2018-07-02
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 112-117 )