An approach on the degree in artificial intelligence applications to model-based diagnosis for dynamic nonlinear CSTR processes is proposed. Stress is placed on residual generation and residual evaluation employing fuzzy logic. Especially for residual generation, a novel Extended Kalman Filter concept, on knowledge Estimator, is introduced. An intelligent fuzzy approach for residual generation and evaluation is also defined. A fuzzy-reasoning approach is proposed to guarantee robustness to the inherent doubtfulness in the identified trends and to provide compact mapping. A two-staged strategy is employed: (i) identifying the most likely fault candidates based on a resemblance measure between the observed trends and the event-signatures in the cognition-base and, (ii) estimation of the fault magnitude. The fuzzy-cognition-base consists of a set of physically explainable if/then rules providing physical insight into the process. This technique provides multivariate differencing and is transparent for fault detection. © 2013 IEEE.