Model based approaches for engine fault diagnosis mostly address the faults external to cylinder since they predominantly use simplified averaged models which do not capture within cycle dynamics. Hence, by using an instantaneous engine model which distinctly characterizes the cylinder’s modes, the events occurring within the cycle can be captured. The events happening across various modes and the engine subsystems can be due to normal operation or faults whose symptoms can be seen as features. In this work, which involves detection and classification of faults occurring in cylinders, is carried out in simulation environment, where, a Kalman filter for state estimation incorporating a nominal instantaneous mode based engine model is considered. Using this estimator as base, faults occurring repetitively (every cycle) are addressed whose features are seen across relevant modes of a cycle. The combination of features of variables such as inputs, states and residuals from a faulty case data are inspected for various engine subsystems across modes within the cylinder’s cycle to detect the faulty cylinder and classify its associated fault.