A Novel Engine Safety State Assessment Model with Two-Layer Belief Rule Base Structure

Assessing the safety status of an engine is critical to the stable operation of electromechanical equipment. The belief rule base (BRB) model enables the integration of expert knowledge to assess the engine safety status, reduce the risk of accidents, and minimize major losses. However, when too many attributes are involved, the rapid growth in the number of rules not only increases the computational cost but also affects the model performance. For this reason, this study proposes a unique interval hierarchical BRB (IH-BRB) modeling approach for engine safety state assessment. The first layer utilizes the interval structure BRB as a classifier, effectively addressing the issue of exploding rule combinations. The second layer classifier employs multiple auxiliary BRBs to transform the multi-classification problem into a binary classification problem for each auxiliary BRB. This approach resolves the challenge of accurately evaluating interval BRB when data similarity is low. To complete the final safety state assessment, interpretability constraints are incorporated into the optimization process. This ensures that the optimized parameters align with expert knowledge and the actual situation. The efficacy of the model in engine safety state assessment is validated using the WD615 engine dataset. Comparison trials and experiments on other datasets further illustrate the advantages of the model in terms of assessment accuracy and interpretability.