Fault diagnosis of diesel engine based on fusion distance calculation

The fault diagnosis of diesel engine is a typical cluster analysis. Effective clustering is often achieved through distance calculation in the cluster analysis. Directing at the issue that fault features cannot be effectively calculated with classic Mahalanobis distance and Euclidean distance when there is fuzzy correlation between the variables of fault feature, this article combines Mahalanobis distance and Euclidean distance to propose a new calculation method with fusion distance. With the method, the correlation coefficients of feature variable are used to determine the weight coefficients for dynamic weighting of Mahalanobis distance and Euclidean distance. The method considers the correlation and independence between feature variables and can effectively improve the accuracy of fault diagnosis. Finally, the effectiveness of the fusion distance calculation is verified from the aspects of diagnosis accuracy and cluster effect with simulation examples.