Fault Diagnosis with SVM and Improved AHP-DS Evidence Fusion
-
Graphical Abstract
-
Abstract
The improved analytic hierarchy process and Dempster-Shafer evidence theory are jointly used to achieve fault diagnosis when evidences are in high conflict. The parameters of the Kernel function of the multi classification SVM and the penalty factor are globally optimized. The fault diagnosis is conducted based on three evidence bodies from SVM. The grading standards for differences, conflicts and uncertainties of the evidence bodies are defined and the hierarchical structure model is built. The AHP is improved to avoid impact of subjective judgment on the construction of comparison matrixes. The credibility weights for the evidence bodies are decided and fusing of evidence bodies are performed. The process is carried out to diagnose the fault of a ship air conditioning system.
-
-