System-Based Component-Test Plans and Operating Characteristics: Binomial Data

Optimum combined test plans for systems and components

“…However, they are usually not system-based, i.e., they are not explicitly designed with system reliability estimation or demonstration in mind. As a result, the actual protection oered by these plans in making inferences about system reliability is not the same as what is advertised; this point is further discussed by Easterling et al [3]. A signi®cant proportion of research in the area of component testing has concentrated on the design of statistically valid, system-based component test plans that also have minimum cost.…”

“…Here we con®ne ourselves to an ad-hoc class of critical regions constructed according to the``sum rule'' which operates as follows: if the total number of failures 1 2 Á Á Á n does not exceed some integer m the system is accepted, otherwise the system is rejected. This rule has been used in the past for series systems (e.g., Altinel [13], Easterling et al [3], and Rajgopal and Mazumdar [10]). It is assumed that system failures are not speci®cally classi®ed as to whether they are caused by component failures or by interface failures.…”

“…Gal [4] appears to have been the ®rst to write on this topic. He was followed by others including Mazumdar [5,6], Yan and Mazumdar [7,8], Rajgopal and Mazumdar [9±10,11], Easterling et al [3], and Altinel [12,13]. All of these papers demonstrate how component tests should be designed from a system perspective.…”

Minimum cost component test plans for evaluating reliability of a highly reliable parallel system

“…For example, a common approach in practice is to simply ''allocate'' system reliability to the individual components comprising the system and then use decision rules for acceptance based on independent tests on these components without any further reference to the system under consideration. As demonstrated in Easterling et al [4], this can lead to very misleading conclusions with risks that are vastly different from the required values. Therefore, component test plans must be carefully designed with the system level objective in mind.…”

Agile manufacturing from a statistical perspective

“…Some €iPnited number: of system tests are necessary for checking these assumptions. Further, the reverse problem of deciding on a suite of component test plans that will yield a system reliability prediction with predetermined precision has also been addressed (Easterling, et al, 1991). In reliability, though, the type of data, e.g., binary padfail data, and system model considered, typically sums and products of component failure probabilities, are apt to be simpler than the process characterization data and process and product models envisioned here, so extending the reliability analogy to the agile manufacturing context will require considerable effort.…”