Clustering-based Sensor Placement for Thermal Fault Diagnostics in Large-Format Batteries

Fault detectability and isolability are essential for realizing online diagnostic algorithms in large format batteries used in safety-critical applications. As sensor locations determine such detectability and isolability, sensor placement becomes a crucial task to enable diagnostics. Limited sensing availability in battery systems makes this issue even more challenging. In this setting, we propose an offline sensor placement framework to maximize the fault detectability and isolability based on limited number of given sensors. Within this framework, we combine physics-based thermal model, fault-to-output transfer functions, and data-driven Evidential C-means (ECM) clustering to determine the essential sensor locations. The performance of the proposed framework is demonstrated via simulation studies on a pouch type battery.