When analysing the fatigue of mechanically stressed units, statistical methods are essential. The tests to investigate the service strength have to be evaluated by statistical methods. In doing so an optimal planning of those test is unavoidable. Many questions arise:

• How to model the loads?
• How to transform these loads into test tracks or test stands?
• Which risk of failure has to be verified?
• Should we use many short or a few long tests?
• How do we extract the the maximum of information from a minimum of tests?