Double Sample Data Fusion Method Based on Combination Rules

This paper proposes a double sample data fusion method based on combination rules to improve the classification of dimensionless indices in petrochemical rotating machinery equipment. This method first collects the original data and counts the mutual dimensionless index as the body of evidence. The reliability of the body of evidence is then determined using a distance calculation method. Finally, the evidence reasoning method is used to fuse the mutual dimensionless index data based on reliability, and the type of fault is detected using the K-S test. A real-time data collection experiment shows that this method can identify the fault type for mutual dimensionless indices that have the appearance of coincidences or evidence conflicts. The experimental results also show that this method has a stronger ability to diagnose faults when compared with the K-nearest neighbor method, and exhibits an accuracy improvement of 9.45%.