In the present study, an extended Kalman filter (EKF) based estimator for a complex chemical process, namely, the Amoco model IV fluid catalytic cracking unit (FCCU), is investigated. This model is multivariable, strongly interacting, and highly nonlinear and is represented by an index one differential and algebraic equation system. The EKF has been modified to be able to handle algebraic state variables. A heuristic using pseudomeasurements is presented to reduce the model linearization errors in the EKF implementation. The performance of the estimator when applied to the Amoco model IV FCCU case study is presented in terms of early fault detection, speed, and estimation accuracy. The results show that, by introduction of pseudomeasurements, the accuracy and robustness of the estimator are improved significantly.