Neural-Networks-Based Fault Detection and Accommodation in a Chemical Reactor

Adaptive Neural-based Fault Tolerant Control for Nonlinear Systems

“…In the event of a failure the control law in normal conditions is augmented with an additional term, which takes into account the on-line approximator and a term to assure the system's stability. Saludes and Fuente (1999) use a predictive controller with a non-linear model calculated by a recurrent neural network to control a reactor tank, after that, when a fault in a sensor is detected, they substitute the fault measurement by the output of the recurrent neural network. Finally, Yu et al (2005) use a multi-layer perceptron network as the process model which is adapted on-line using the extended Kalman filter to learn changes in the process dynamics.…”