An Interpretable Belief Rule-Based Fault Diagnosis Method for Complex Equipment Considering Linguistic Fuzzy...

Area of Science:

• Engineering
• Artificial Intelligence
• Information Science

Background:

• Linguistic fuzziness and cognitive variability pose challenges in fault diagnosis.
• Existing methods risk model distortion during optimization.
• Expert knowledge is often imprecise, leading to uncertainty in diagnostics.

Purpose of the Study:

• To propose an interpretable belief rule-based fault diagnosis method for complex equipment.
• To address challenges of linguistic fuzzy information and expert cognitive variability.
• To ensure model usability and interpretability in practical engineering applications.

Main Methods:

• Constructed an interval grey number table to convert linguistic fuzzy information into interval grey representations.
• Introduced a certainty degree fusion method with fusion weights to mitigate low-confidence evidence.
• Designed an interpretable parameter optimization method with dynamic knowledge anchoring.

Main Results:

• The proposed method quantifies uncertainty in expert judgments while preserving knowledge boundary information.
• Fusion weights mitigate interference from low-confidence evidence during rule generation.
• Dynamic knowledge anchoring constrains model parameters within expert-defined bounds.

Conclusions:

• The developed method achieves superior diagnostic performance for complex equipment.
• The approach ensures model usability and interpretability in practical engineering scenarios.
• It effectively handles linguistic fuzzy information and expert cognitive variability in fault diagnosis.