Since data fusion in the process of the traditional fault diagnosis method is not accurate enough, it is difficult to use the dimensionless index to distinguish among fault types of… Click to show full abstract
Since data fusion in the process of the traditional fault diagnosis method is not accurate enough, it is difficult to use the dimensionless index to distinguish among fault types of problems. This paper proposes a data fusion method based on mutual dimensionless. This method uses real-time acquisition of original data and dimensionless calculations, obtains five dimensionless indices for each dataset, and then uses support vector machine (SVM) model projections for the dataset to judge fault types. Using dimensionless indices to process raw data, the SVM method for training can more effectively solve the problem due to the imperfection of the old dimensionless index leading to a low accuracy of fault diagnosis. Using a petrochemical rotary machinery experiment, the accuracy of the method of fault diagnosis is higher; in a single experiment, the fault detection accuracy can reach 100%, where compared with the traditional dimensionless index data fusion method, the accuracy is increased by 20.74%. The method has stronger ability to judge failures.
               
Click one of the above tabs to view related content.