AbstractKnowledge of the manner in which machines break down is a vital component in any simulation model. It is this information that is used to interrupt the cycling of the model and to simulate the interaction of machine downtime. Usually, of all the parameters such as toolchangc and manning, the downtime has the greatest effect on the model throughput. It is essential, therefore, that the method simulation models used to simulate breakdown behavior are made as accurate as possible. This paper describes the results of a four-month study on 10 transfer line machines. The machines perform the first few machining operations on an engine cylinder block line. Information relating to the breakdown behavior was collected and tested for goodness of fit with standard statistical distributions. As a further comparison two simulation models of the 10 machines were built. Breakdowns on the first model were simulated by sampling standard distributions, and on the second model they were simulated using actual historical data. This comparison verified that the use of simulated breakdowns produed results as accurate as those found using historical data.