Improved Fault Isolability for Solid Oxide Fuel Cell Diagnosis Through Sub-system Analysis

This work illustrates an advanced technique able to improve fault isolability. Univocal fault isolation may be hindered by on-board sensor reduction, since fault clustering can occur. According to the proposed technique, isolated system component models, fed with faulty inputs, are used to compute a set of redundant residuals reacting only if the related component is under faulty state. The technique is characterized and tested via simulation analysis, on a Solid Oxide Fuel Cell Anode Off-Gas Recycling system. Stack internal leakage and fuel heat exchanger crossover models are implemented in the complete system model and used to simulate both nominal and faulty conditions. The achieved results confirm the capability of the proposed approach to univocally isolate the considered faults.