Efficient On-line Fault Isolation and Identification in TRANSCEND for nonlinear systems

Integrated System Health Management methodologies require efficient fault identification approaches to ease system reconfiguration or to adapt controllers to accommodate for the faulty behaviors. However, accurate and timely on-line fault identification of complex, nonlinear systems is difficult and can be computationally expensive. In this work, we extend the quantitative fault isolation and identification proposal in TRANSCEND. We propose a more efficient approach based on previous works on model decomposition and dependency compilation. We generate a set of smaller estimation tasks from the global estimation problem to reduce the computational burden. We tested the approach in a nonlinear three tank system. Current results show a considerable reduction in fault identification time, without compromising the accuracy in the estimation.