Health factor datasets extracted from clean and sensor-fault-injected NASA Li-ion battery data

This repository contains the health factor (HF) datasets used in the study on uncertainty-driven model selection for fault-prone SOH prediction. Five voltage-, energy-, and temperature-derived health factors (HF1–HF5) were extracted from the NASA Li-ion battery degradation dataset (cells No. 5, 6, 7, and 18) under both clean and voltage-sensor-fault-corrupted conditions. Four sub-threshold sensor fault types — bias, drift, precision noise, and stuck-at — were synthetically injected at three severity levels (low, medium, high), applied both individually and in combination, to replicate realistic, undetectable BMS sensing degradation.