Adaptive model-based active tolerant control for nonlinear process

A new Fault Tolerant Control (FTC) scheme based on the inversion of adaptive RBF neural network model for unknown multi-variable dynamic systems was proposed. An adaptive RBF model was designed to build process model and was adapted on-line by using Extend Kalman Filter (EKF) technique to learn fault dynamics caused by component faults. Then, an inversion of the RBF model was used to the adaptive controller to maintain the system performances after fault occurrence. The proposed scheme was applied to a multiple variable three-tank process to demonstrate the performance of the scheme.