Industrial Process Approach to Fault Diagnosis and classification ds evidence Fusion

The present invention discloses a process industrial fault diagnosis method based on multiple classifiers and DS evidence combination, the method first independently resampled data according to the failure of an industrial process, a plurality of classifiers and the new method for the application of training data to obtain offline modeling each model, while the performance of each classifier fusion method shown in the form of a matrix. Then, basic probability calculated by different categories of evidence theory DS assignment function, according to the similarity index of multiple classifiers integration and decision-selective comprehensive, joint obtain basic probability assignment function, by comparison, to get the final results of the diagnostic classification . Compared to other existing methods, the present invention can greatly improve the industrial process diagnostic effect, reduce the delay time of diagnosis, improve the diagnostic accuracy, performance monitoring largely improved, enhancing the process operator of the process operation and understanding confident, more conducive to industrial process automation implementation.