Abstract :
The note provides some direct calculations on which to base a control procedure for the reliability of systems that can be built up from various arrangements of series and parallel subsystems. Its application is intended to situations where a fully assembled system can not be tested as such, and where there is available only attribute test data on the components of the system. The reliability of an arbitrary system is often representable as some mathematical function of the form F (p1, p2, ... pn) where the p´s denote binomial parameters or success probabilities associated with the components of the system. We take as observables the corresponding p¿´s representing attribute test data on the components (we are considering situations in which fully assembled system tests are precluded). We then consider the associated function of random variables F¿ (p¿1, p¿2, ... p¿n) for certain more or less typical systems, that is, systems built up from various series and parallel arrangements of components. The control procedure is then based on some directly computed statistical properties of F¿.
