A Causality-Informed Correlation-Aware Health-State Assessment for Complex Equipment.

Area of Science:

• Engineering
• Data Science
• Systems Science

Background:

• Health-state assessment is crucial for prognostics and health management (PHM) of complex equipment.
• Existing methods often neglect statistical dependence from causal coupling, leading to assessment bias.

Purpose of the Study:

• To propose a novel health-state assessment model, the evidential reasoning rule considering causality-informed correlation (ERr-CIC).
• To address limitations in current PHM techniques by incorporating causality-informed correlation.

Main Methods:

• Analysis of causal coupling relationships and their impact on health assessment.
• Application of convergent cross-mapping (CCM) to examine subsystem causal coupling.
• Development of the ERr-CIC model using a discount factor, conditionally hybrid correlation coefficient (CHCC), and signaling sequences for fusion order.
• Sensitivity and robustness analysis of the model to CHCC.

Main Results:

• The ERr-CIC model effectively quantifies causality-informed correlation using CHCC.
• Sensitivity analysis identified key parameters and confirmed model reliability under perturbations.
• Experimental validation on the PAMD simulation device demonstrated competitive accuracy, stability, and interpretability.

Conclusions:

• The proposed ERr-CIC model offers a more accurate and reliable approach to health-state assessment in complex equipment.
• Accounting for causality-informed correlation is essential for mitigating bias in PHM.
• The model provides a balanced performance in stability, interpretability, and accuracy.