THE TRINOMIAL FAILURE RATE MODEL FOR TREATING COMMON-MODE FAILURES

Abstract Common cause, or common mode failures are a major concern in analyzing system reliability. This paper presents a model, called the Trinomial Failure Rate model, that allows careful and elaborate treatment of common cause failures. To date, the models that have been used as tools for common cause failure analysis are of the binary type, that is, a component state is either a success or a failure in describing component failure events. The simple on-off method cannot explicitly analyze existing complex phenomena such as partial failures, incipient failures and potential failures. The trinomial Failure Rate model has been developed to explicitly analyze such ambiguous events. The maximum likehood estimation method is used to develop estimators of model parameters. The results of the Trinomial Failure Rate model are compared with those of the Binomial Failure Rate and the beta-factor models using assumed data for the events that fall into the gray area between success and failure. By identifying the ambiguous situations and subdividing the events which are used as input data, events that cannot be clearly classified as success or failure can be discovered so that more detailed system analyses can be performed.