On the limitations of redundancies in the improvement of system reliability.

Résumé

Some program managers share a common belief that adding a redundant component to a system reduces the probability of failure by half. This is true only if the failures of the redundant components are independent events, which is rarely the case. For example, the redundant components may be subjected to the same external loads. There is, however, in general a decrease in the failure probability of the system. Nonetheless, the redundant element comes at a cost, even if it is less than that of developing the first one when both are based on the same design. Identical parts save the most in terms of design costs, but are subjected to common failure modes from possible design errors that limit the effectiveness of the redundancy. In the development of critical systems, managers thus need to decide if the costs of a parallel system are justified by the increase in the system's reliability. NASA, for example, has used redundant spacecraft to increase the chances of mission success, which worked well in the cases of the Viking and Voyager missions. These two successes, however, do not guarantee future ones. We present here a risk analysis framework accounting for dependencies to support the decision to launch at the same time a twin mission of identical spacecraft, given incremental costs and risk-reduction benefits of the second one. We illustrate this analytical approach with the case of the Mars Exploration Rovers launched by NASA in 2003, for which we had performed this assessment in 2001.