Battery UQ & XAI Digital Twin (IEEE Xplore 11249263)
01: McMaster Battery Dataset (Train)
02: Oxford Battery Dataset (Cross-Test)
Back to Portfolio ↗
← Back to Curve
McMaster (Train)
Oxford (Cross-Test)
90% Band
95% IEEE Target
99% Safety
Degradation & UQ Curve
4-Method XAI Attribution
Full Screen ↗
Empirical Coverage (PICP)
95.2% Verified
Calibration Error (ECE)
0.012 ECE
Interval Width (Epistemic + Aleatoric)
+/- 1.8% SOH
Top XAI Driver (SHAP R^2 = 0.94)
Terminal Voltage Drop (dV/dt)
BUBBLE 1 OF 4
×
Cycle 120: Early Healthy Regime
CNN Prediction: 89.2% SOH | 95% Interval: [87.4%, 91.0%]
Lithium-plating and solid-electrolyte interphase growth accelerate capacity fade past Cycle 340. Click "XAI Drivers ->" below any node to inspect the 4-method feature attribution breakdown.
Open 4-Method XAI Attribution →