A modeling method termed as the discrete time hybrid Diagnosis (DTHD) is presented to constructure models for a hydraulic punching machine. The DTHD method is an extension of the works on FDD (the fault detection and diagnosis) which are based on FSM-based modeling. FDD of DTHD models are suited for real time systems because of their capability of detecting faults leading to failures in terms of erroneous delays and deadlines. The occurrences of faults or failures are represented for individual system components by categorizing the set of states of each individual system component model into normal or failed conditions of various types. The condition corresponding to the states (of a component model) are represented by an unmeasurable variable. The components are then composed to produce the overall representation of failures on the whole system model with a single unmeasurable variable.