Sometimes the reliability of a system is demonstrated by means of testing procedures. For example, a proof load is applied to the system of interest. If the system successfully withstands the applied load, it is considered sufficiently safe. The higher the target reliability that must be maintained, the larger the applied proof load has to be. However, the danger that is associated with higher proof loads is the increased potential of failures of the system due to the applied testing procedure.
The monetary costs of failure during proof testing are typically small compared to system failure during operation. Nevertheless, the costs of failure during proof testing can still be considerable, as the tested system is lost or at least partially damaged. One strategy to minimize the total costs is to select the least demanding proof load that still maintains the targeted reliability. Another strategy is to demonstrate that the targeted reliability level is maintained without conducting an expensive proof load test. Both strategies require an adequate probabilistic model of the system of interest.
We assisted our clients in setting up probabilistic models to describe the uncertainties in their systems of interest. Based on this, we evaluated the reliability of their systems as a function of the employed testing procedures and the applied proof load. In a next step, we took the experience and data gathered during all already conducted tests into account in order to reduce the uncertainties in the underlying probabilistic model. Finally, we assessed whether it is economical and justifiable to omit the expensive proof testing procedures.