Inheritable Object-Oriented Bayesian Networks: An Efficient Fault Diagnosis Model for Complex Equipment Systems

Bayesian network is a widely utilized model for equipment fault diagnosis. Nonetheless, when facing complex systems, the construction and inference of BN gets more difficult. Inspired by object-oriented programming principles, the paper proposes a fault diagnosis method based on Inheritable Object-Oriented Bayesian Networks. The core idea is to define similar segments in Bayesian network as classes and build the model by instantiating these classes. On this basis, knowledge reuse and model extension are realized through class inheritance. In addition, this paper proposes a parallel inference method to reduce the computational complexity of diagnostic reasoning. The effectiveness of the modeling and inference is demonstrated through an engine fault diagnosis case.