Health state assessment method for complex system based on multiexpert joint belief rule base

Area of Science:

• Engineering
• Artificial Intelligence
• System Health Management

Background:

• Complex systems degrade over time, necessitating accurate health state assessment.
• Belief Rule Base (BRB) is a valuable tool for managing uncertainty in health assessment but faces challenges with expert knowledge integration.
• Inconsistent expert cognition and incomplete knowledge hinder traditional BRB model construction and interpretability.

Purpose of the Study:

• To propose a novel health state assessment method for complex systems using a joint multiexpert belief rule base (BRB-ME).
• To address limitations in expert knowledge completeness and cognitive consistency in BRB modeling.
• To enhance the accuracy and stability of complex system health assessments.

Main Methods:

• Experts individually construct their BRB models.
• A novel multiexpert knowledge fusion algorithm is developed to integrate diverse expert models.
• The Evidential Reasoning (ER) approach serves as the inference engine.
• A constrained multi-population evolutionary whale optimization algorithm (C-MEWOA) is employed for optimizing the BRB-ME model.

Main Results:

• The BRB-ME model successfully fuses knowledge from multiple experts.
• Case studies on lithium-ion batteries and flywheels demonstrate the model's effectiveness.
• Comparative analyses confirm the BRB-ME model's superior stability and accuracy in health state assessment.

Conclusions:

• The proposed BRB-ME method effectively integrates multiexpert knowledge for complex system health assessment.
• The approach overcomes limitations of traditional BRB methods regarding expert knowledge.
• BRB-ME offers significant improvements in assessment stability and accuracy, validated by practical applications.