Prognosis of material and product properties

In many complex systems and processes, it is often completely unclear at the beginning on which potential influence factors a selected performance parameter depends. This is due to a lack of adequate physical models. In particular, the known dependencies are frequently nonlinear, and vary with the state of the dynamical system.

If, however, sufficient representative data is available, e.g., stemming from systematic series of experiments characterizing the input-output behaviour, a system description in the form of a black box or grey box model can be derived by appropriate methods from system identification, data mining, and statistics. These models can then be used for the task of prognosis, especially allowing for the derivation of system sensitivities with respect to selected influence parameters in their particular definition ranges.

These methods therefore are an effective tool in the hands of engineers whenever they have to decide which input variables must be changed by which order of magnitude, in order to yield a higher probability for the task variable to behave in the desired way under the considered loading scenarios.