This paper presents a method for detecting abrupt changes in dynamic systems. It is based on statistical information generated by the Extended Kalman Filter and is intended to reveal any drift from the normal behaviour of the process. This method was originally developed for linear systems (Himmeblau, 1978), our contribution consists in extending it to unsteady state nonlinear systems. A failure of chemical origin in a perfectly stirred batch chemical reactor, occurring at all unknown instant, is simulated. The reactor is modelled by mathematical expressions translating the relationships between heat and mass balances applied to the reaction mass. These expressions constitute the model which describes the behaviour of the process under normal conditions. The Purpose is to detect the presence of this abrupt change, and pinpoint the moment it occurred. It is also shown that the convergence of the Extended Kalman Filter is accomplished more or less rapidly according to the nature of the noise generated by the measurement sensors. The state estimate is observed and discussed, as well as the delay in detection according to the decision threshold. The reaction under Study is an exothermic second order one.