THE GENERATION AND USE OF COMPONENT FAILURE RATE DATA

This paper was presented at a symposium held in Sydney in November 1976 by the New South Wales Branch of the Statistical Society of Australia. The accuracy with which system or equipment reliability can be predicted depends on the component failure-rate data used. Whatever the source of the data, they are suspect; the selection of the correct data is as much a matter of subjective judgement as of objective reference to handbooks or data files. The reasons for this state of affairs are many. The mortality of a component is a function of a number of variables in the design and manufacture of the equipment as a whole as well as of the component itself, and is also a function of variables in the operational and environmental stresses. The mortality is further affected by the maintenance strategy and environment. Moreover, uncertainities arise from possible data obsolescence, from the combined effects of simultaneous field stresses, and from doubt as to whether failure rates are constant with time or whether a failure-time distribution other than the negative exponential applies. The Author discusses how to reduce these uncertainities, and provides guidelines for the generation, processing, interpretation, and use of component failure data. It is stressed that the assumption of constant failure-rate, largely accepted in electronic-equipment design, always requires attention, particularly when dealing with mechanical components. However, inaccuracies (from various causes) in failure-rate figures often overshadow any errors due to an invalid assumption of exponential failure-time distribution; it is only in the region of early "infant" mortality or when close to wear-out that the distribution discrepancy becomes the main source of error in equipment reliability prediction. Order from: BSRA as No. 47,505.