.png){kind=icon}
.png){kind=icon}
Return
Interpretable Layered Health Assessment for Gas-Insulated Switchgear via Differentiated Thresholding and Dempster–Shafer Fusion
.png){kind=icon}
PRE
AI
C
H
P
C
H
B
DOI：10.1109/tii.2026.3684102.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Gas-insulated switchgear (GIS) condition assessment must cope with heterogeneous components and monitoring variables while remaining interpretable for maintenance decisions. This article proposes an interpretable layered health-assessment framework that couples reliability-aware differentiated thresholding with Dempster–Shafer (D-S) evidence fusion. Historical monitoring records are fitted with a Weibull model to derive reliability-related priors, and differentiated thresholds are estimated by integrating guideline limits, voltage-category subgroup statistics, and sample-size-weighted fusion. A lightweight neural calibrator is used only during offline threshold learning to refine voltage-specific boundaries, while online inference stays rule-based. Four-layer architecture (parameter-performance-component-equipment) maps observations to fuzzy memberships and basic belief assignments (BBA), applies degradation-aware analytic hierarchy process weighting, and performs hierarchical D-S fusion to output four ordered health grades (normal/attention/abnormal/severe) with cross-layer explanations. Experiments on a 220 kV GIS insulator-defect platform (184 tests) achieve 95.65% accuracy. Fleet-scale validation on 621 GIS units achieves 95.27% accuracy, improving accuracy by 8.23% over guideline-based assessment, reducing false positive and false negative rates by 6.30% and 12.00%, and increasing recall by 12.00%. The framework models uncertainty under missing indicators via BBA and D-S fusion and decouples offline learning from online assessment, achieving 2.02 ms per-sample inference on a single-thread central processing unit.
Keywords:
Condition assessment
Dempster—Shafer (D-S) evidence fusion
differentiated thresholds
fuzzy membership
gas insulated switchgear (GIS)
Journal
IF:
9.9
Papers:
8.3K
Citations:
6.0W
Organization
Cited Papers