BURN-IN VS. INFANT MORTALITY

Reliability theory deals with the reliability of components and the systems formed by them; closely allied is maintenance theory, which devises optimum strategies of keeping such systems working under given requirements (usually at minimum cost).

Reliability itself is a concept that depends on time: it is the pro-bability that a component or system will operate from the moment it is put into service to a time t. Thus reliability is a curve that drops from 1 at time 0 (initiation) to zero at time infinity, for nothing works forever. The reliability of a system is calculated from the re-liabilities of its subsystems and components: by configuring them in certain ways (especially by including redundant components that take over as others fail), one can increase the system reliability high above that of its individual components.

In burn-in, the defective light bulbs (say) in a sample of thou-sands pop frequently, and the failure rate is high. These failures withdraw the bad bulbs from the sample, so that the failure rate of the remaining bulbs decreases and settles at a low (quite often constant) value. But when the filaments wear thin, they, too, start popping and the period of wear-out begins.

It may seem cynical to compare this with human life, where the analogous periods are infancy, maturity and old age. However, the basics of reliability theory were indeed modeled after vital statistics, and mathematical symbols do not care what they stand for. Moreover, the connection is not merely formal. A child burdened by inherited disease, or by insufficient resistance to the stresses of its environment, or by other defects, is more likely to die early in its life, just as a faulty component is more likely to fail earlier than later, and that is what causes the bulge in the mortality curve, known as infant mortality in one case and as burn-in in the other.

Similarly, the causes of death or failure during the main part of life are mostly accidental or spurious; and the actual decay of the organism or the component sets in with old age or wear-out.

From these fundamental measured curves, one can use pro-bability theory to derive the reliability of a system, which in turn can be used to calculate the optimum maintenance strategy to keep the system operating at minimum cost (or alternatively, at minimum down-time for repairs, or possibly at a specified com-promise between the two).