Assessment and modelling of baseline design under consideration of stochastic properties

Cost optimal design minimizing sum of production and failure cost (© 2014 DYNARDO Austria GmbH)
Work flow leading from random uncertainty of input data to stochastic description of stack performance (© 2014 DYNARDO Austria GmbH)


Work Package Objective

The work package synthesizes physical models of structural, thermal, electrical, and electro-chemical behavior of stacks into a simplified meta-model to be used in the probabilistic analysis as fast surrogate solver. This meta-model consists of simple mathematical functions (e.g. polynomials) relating the input parameters (e.g. strength) to specific output parameters (e.g. number of cycles to failure). Based on these meta-models, the design of the stack can be optimized with respect to both performance and cost in a computationally efficient way.


Specific Challenge

The meta-model obtained in this way allows the prediction of the output parameters for combinations of the input parameters not covered by full model simulations. This is needed to obtain stable estimates for the failure probability, which should be rather small in the final stack design. As the meta-model must be adjusted to represent the output quantitatively correct for different values of the input (e.g. to cover the inherent stochastic variability), it is essential to know the range of variability of the input beforehand. Another complicating factor is that many input parameters cannot be described by single values (random variables), but are spatially varying with significant spatial correlation (random field).


Work Package Leader:

Organisation: DYNARDO Austria GmbH

Name: Christian Bucher