A complex system health state assessment method with reference value optimization for interpretable BRB.

Area of Science:

• Engineering
• Artificial Intelligence
• System Science

Background:

• Health condition assessment is vital for complex system maintenance and safe operation.
• Existing belief rule base (BRB) models offer interpretability but suffer from parameter inaccuracies and reduced transparency post-optimization.
• Accurate assessment requires handling uncertainties and maintaining interpretable, traceable results.

Purpose of the Study:

• To propose a novel interpretable belief rule base with reference value optimization (I-BRB) for complex system health condition assessment.
• To address inaccuracies in expert-provided parameters and maintain interpretability during model optimization.
• To enhance the accuracy and reliability of health condition assessment models.

Main Methods:

• Developed a reference value optimization algorithm with interpretability constraints to refine expert-provided parameters without compromising knowledge.
• Employed the projection covariance matrix adaptation evolution strategy (P-CMA-ES) with interpretability constraints for optimizing remaining parameters.
• Validated the I-BRB method through a case study on bearing components of a flywheel system.

Main Results:

• The proposed I-BRB method effectively optimizes reference values while preserving expert knowledge.
• Parameter optimization using P-CMA-ES improved model accuracy without sacrificing interpretability.
• The case study demonstrated superior accuracy of I-BRB in health condition assessment compared to existing methods.

Conclusions:

• The I-BRB method provides a robust and interpretable approach for complex system health condition assessment.
• This approach enhances accuracy and maintains transparency, crucial for optimizing maintenance strategies.
• I-BRB offers a promising solution for reliable system monitoring and diagnostics.