Subproject A8

In the second funding period, stochastic uncertainty in the development phase based on fixed models of the load-bearing system was the central topic. In contrast, in the third funding period it will be allowed that the model itself is uncertain. In the usage phase values of the real system will be observed and used to improve the model. Essential questions in this context are: How can model uncertainty be identified? How can the model be improved in case of model uncertainty? How can model uncertainty be considered in the evaluation of uncertainty?

Concrete Work Program

• Identification of Model Uncertainty:
  Development and analysis of methods, which are able to consider changes, caused by model uncertainty, in the prediction of stochastic uncertainty by density respectively quantile estimation.
• Improvement of Models in Case of Model Uncertainty:
  Development and analysis of methods to improve the model based on real values of the load-bearing mechanical systems including methods to select parameters automatically. The developed methods will be analyzed theoretically with regard to consistency, rate of convergence and adaption. In addition, the finite sampe size behavior of the methods will be analyzed using simulated data.
• Evaluation of Stochastic Uncertainty in Case of Model Uncertainty:
  Development of methods to predict stochastic uncertainty by estimating confidence regions. Simulated data will be used to illustrate the method.

The new developed methods will be applied to the SFB-Demonstrator.