System Reliability

Stochastic sensitivity analysis and statistical experimental design

In the real world there are many systems whose behaviour cannot be easily predicted. Even for the systems which can be simulated on a computer, a change in the parameters can have unexpected consequences. Interactions between the components of a system and their influences must be known beforehand for many applications. Therefore, in the research group SAM, research is being carried out to find new methods of sensitivity analysis for both experimental and numerical applications to detect and influence main effects and interactions. Recent research works are:

• the implementation of different methods for global sensitivity analysis of complex systems,

• the improvement and exploitation of different potential sensitivity analysis,

• the numerical simulation and analysis of different methods on academic reference systems,

• the evaluation of the convergence behaviour of the methods and the quality of the results,

• the qualitative experimental validation of the results of the numerical simulations by use of statistical design of experiments,

• the derivation of a general methodology for sensitivity analysis for complicated and complex systems and their transfer to industry-oriented systems.