Abstract :
The time dependence of the failure rate of electronic parts and equipments has been strongly evidenced by recent data. Information is cited which reveals that for many devices failure rate not only is time dependent but is an inverse time function, i.e., decreases as time increases. The possibility that this property is inherent in many other parts and systems is hypothesized. The electronics industry will be confronted with new and perplexing problems when the non-constant failure rate is treated with realism. Reliability estimations (predictions) in the proposal stages for example, will involve new concepts in failure rate numbers which are no longer "F", but "F(t)". It is anticipated that these functions will be unwieldy. Useful data will be scarce for some time to come. This industry must develop improved reliability estimating and testing techniques to realize the benefits of this more accurate mathematical function, F(t). An important added capability, that of part improvement through long burn-in, will accrue for parts that are characterized by decreasing failure rates. Customers, in requiring part and subsystem suppliers to prove failure rates by test, in accord with Darnell Report2and AGREE Report1philosophies, should insist that the time dependence of failure rate be established. Industry-wide and government coordinated efforts appear necessary if we are to achieve data that will enable us to take our long-buried heads out of the sand.
