Research On Problems Of Pumping-jack Diagnosis Based On Support Vector Machine

Support vector machine is to solve problems in machine learning by means of optimization methods. It has unique advantages in addressing the small sample data sets and nonlinear problems. It is able to resolve the problem in pumping-jack failure for such a small sample. Support vector method is applied to the fault diagnosis of oil pumping-jack in the paper. Specifically completed the following work:In the paper, common method for intelligent fault diagnosis and limitations is performed a detailed analysis and study, and depth research is done on common solutions for this small sample of pumping failure problem. It does research on support vector machine. Multi-class classification problems are discussed, it focuses on analysis of the "one-against-one" and "aodnape-teadg.a iTnhste- roersitg."i n aTl hkee r nkeelr nfueln c ftuionnc tiios nm iosd iifmiepdr obvye tdh. e Fsiarmstlpyl,e addaatap,t iavned ktehren eclo r fruenspcotinodni n igs kernel function is get. It can improve accuracy and classification speed, and reduce the number of support vectors to a certain extent. Secondly, a mixed kernel function based on an improved Gauss kernel function is proposed. The kernel functions have the performance both of the single Gauss kernel function and of polynomial kernel function. The kernel functions have the advantage of global learning ability and generalization ability.Support vector machines are used in oil pumping-jack fault diagnosis; diagnostic results are compared, and impact of relative parameters on the diagnosis results is discussed.