Based On The Fuzzy Support Vector Machine Hydraulic System Fault Diagnosis

As the development of national economy and modern industrial technology,hydraulic pressure and hydraulic transmission technology is applied inengineering machinery widely. Hydraulic system failure information has certainfuzziness,the reason which produces the same phenomenon is not the same onenecessarily. It is difficult to find a specific reason for failure and fault locationwith a single fault phenomenon.In view of common problems of the current intelligent fault diagnosissystem, such as large fault knowledge base, capacity limitations when solve theproblem, poor ability of automatic acquiring knowledge, do not according to theconcrete analysis of faults, fuzzy support vector machine (FSVM) was appliedto the fault diagnosis of hydraulic system which resolve the problem of existingalgorithm ignore or don’t pay attention to the fuzziness of fault data. Existingtwo kinds of fuzzy support vector machine (membership fuzzy support vectormachine model and fuzzy decision-making fuzzy support vector machine) bothhave shortcomings, so proposing the double network model fault diagnosissysterm. The first layer is the main network of sample fuzzy membershipachieves simple classification, the second network using decision-making fuzzysupport vector machine (FSVM) model get a diagnosis according to fuzzyclassification rules. Experiments show that fuzzy support vector machines whichbased fuzzy theory and machine learning algorithm have very good learningability, double network model of fault diagnosis system overcome the highfuzziness of failure data and got high diagnosis precision.In order to improve the performance of the machine learning deeply, thispaper proposing optimization algorithm for original classic sequences minimumalgorithm. The improvement and optimization method based on FSVM trainingalgorithm and parameter selection for. Training algorithm is improved in themain choice and two threshold value of the variable optimization methods wereput forward concrete improvement methods which accelerate the algorithm oflearning and convergence rate. For FSVM parameter selection problem, as the theoretical basis of genetic algorithm and simulated annealing algorithm putingforward a kind of adaptive simulated annealing genetic algorithm which used tooptimize the parameters of learning machine choice. Experiments show thatSVM learning performance is increased significantly after algorithm to optimizethe parameters.After system optimization, fuzzy support vector machines have betterlearning ability and the classification accuracy, suitable for the classification offuzzy sample data. Optimized fault diagnosis system will used in hydraulicsystem fault diagnosis of research and have the practical significance of thepractical and feasible. But because fuzzy support vector machine (SVM) is theartificial intelligence and data mining discipline of the direction of a newresearch area, as a kind of new technology, fuzzy support vector machine (SVM)model still has large research and development space.