The talk gives a tutorial introduction to fault diagnosis of discrete-even tsystems which
are represented by automata. According to the principle of consistency-based diagnosis,
methods for detecting and identifying faults in a dynamical system are based on the
investigation whether the measured sequences of input and output symbols are
consistent with the model of the faultless or the faulty system. As usually no a-priori
information about the initial state of the system is available, the diagnostic algorithm
has to solve a state observation problem, which will be described at the beginning of
the presentation.The basic idea of this method is then used for diagnostic purposes
and later generalised to nondeterministic, stochastic and timed automata. Applications
to industrial processes illustrate the effectiveness of the method.