Defining mean time-to-failure in a particular failure-state for multi-failure-state systems

In multi-failure-state systems which have both dangerous (critical) and saf e (noncritical) failure states, it is both meaningful & important to be abl e to measure the MTTF in a particular failure state separately from MTTF in general. MTTFD (fail dangerous) and MTTFS (fail safe) are defined, emphasi zing the engineering context of the problem. Methods are developed for comp uting these indexes. These methods require only finding the MTTF of a modif ied Markov model. The model modifications are relatively simple and can acc ount for the effects of various repair & maintenance policies. An example i llustrates the methods. These developments are related to the classic theor y of Markov models. Several other methods for computing 'mean time to criti cal failure' are shown to be special cases of the more general method prese nted here.