EFFECT OF DISTRIBUTIONAL ASSUMPTIONS ON SPARE COMPONENT AVAILABILITY EVALUATION
This thesis presents some basic techniques for availability modelling and quantitative evaluation in repairable systems. Two Monte Carlo simulation methods are developed for the evaluation of Markovian systems with spare components. The· results of the simulation method are compared with those obtained by analytical methods. Markov techniques can not be applied directly if the distributions associated with the state residence times are not exponentially distributed. Availability evaluation of on-Markovian systems is usually much more complicated than that of a Markovian system. A practical non-Markovian simulation technique is discussed in this thesis and characteristics of important distributions are presented. A range of studies for selected repairable systems are conducted to study the effect on system unavailability of various distributions. The failure rate is assumed to be constant and distributional variation is applied to the repair and installation processes. In order to facilitate comparison, the mean values associated with the repair and installation times are held constant. The changes in unavailability are therefore due entirely to the changes in the shape of the distributions. The thesis provides a general appreciation of the effect of non-exponential residence times on basic system availability indices.