Optimum reliabilities and optimum design

Abstract The designer's task is to optimize. Designing for an intuitively fixed reliability is almost as arbitrary as designing for an intuitively fixed safety factor. Efforts to improve the calculation of reliabilities of proposed designs are commendable but sterile unless we take the additional step of optimizing the design reliabilities associated with the various potential limit states. Through examples it is shown how these reliabilities depend on the consequences of entrance into limit states and on the marginal initial cost of the reliability and how earthquake design spectra must differ radically from those for fixed return periods. It is further claimed that the optimum reliabilities depend on how a structure enters a limit state. A more realistic aim is not the establishing of design reliabilities but of optimum designs. This is illustrated in connection with progressive collapse. Mistakes (“human” or “gross” errors) are not considered in the present analysis.