Mean Value Engine Models, commonly used for model based fault diagnosis of SI engines, fail to capture the within-cycle dynamics of engines, often resulting in reduced fault sensitivity. This paper presents a new Hybrid Automata based modeling approach for characterizing the within-cycle dynamics of the thermo-fluidic processes in a Spark Ignition Gasoline Engine, targeted for use in model based fault diagnosis. Further, using a hybrid version of the Extended Kalman Filter (EKF), the states from the nonlinear hybrid automata based dynamic model are estimated and their results validated w.r.t standard industrial simulation software, AMESim. It is observed that due to the switching of within cycle engine dynamics, causing mode change, there is a corresponding change in model's structure which in turn can cause change in system's observability. The changes in observability (with modes) and the characterization of the evolving instantaneous engine dynamics into distinct system major modes and sub modes is expected to provide better sensitivity and faster detection and isolation of faults, when estimating the state variables in real time.
