Research on intelligent fault diagnosis method for complex equipment based on decision-level fusion

Rough Bayesian network classifier (RBNC) not only has the ability of rough set (RS) for analyzing and reducing data, but also holds the advantage of Bayesian network (BN) for parallel reasoning, and it has been successfully used in fault diagnosis field. However, single RBNC sometimes will produce misdiagnosis because of many uncertainty factors in practical diagnostic process. In order to improve the accuracy rate and reduce the uncertainty, a new decision-level fusion diagnosis method was proposed based on D-S evidence theory. Firstly, the RNBC models are taken as evidences and the posterior probability values of all fault types as correlation coefficients. Then, all evidences are synthesized by using the combination rule of evidence theory, and consequently, the diagnostic result can be gained by the decision-making method based on the basic probability number. Finally, the validity and engineering practicability of the proposed method is demonstrated by an example of fault diagnosis for diesel engine, and the results show that the proposed method is more effective than the RS method and the BN method and the RBNC method.