PhD Position: Identification of Physical Parameters in the Presence of Model Deficiencies, Delft University of Technology, Netherlands

Classical parameter estimation methods are essentially dependent on correct specifications of the underlying model and knowledge of the statistical properties of the random disturbances in the data. If models are misspecified or approximative, the results of these methods become unreliable: inconsistent parameter estimates result and subsequent inferences are suspect. The presence of model deficiencies can be caused by the fact that exact knowledge of all physical phenomena is still lacking, or by the fact that taking account of all physical phenomena leads to an approach that is too complex to be feasible. In the system identification research of the last decade, appropriate attention has been given to aspects of misspecified models when estimating dynamical models from experimental data. However, the link with physical models and physically interpretable parameters is not yet explored. In this project, an attempt will be made to bridge this gap, aiming at improved (physical) parameter inference in real world applications. The goal of the project is to develop a method that takes account of the fact that the chosen model contains deficiencies, while at the same time the data are noise-corrupted. For this purpose, two sources of uncertainty (model and data) have to be combined in an optimal way. This should lead to physical parameter estimates with a specific quantification of their uncertainty.

Contact persons/Web link :
http://www.dcsc.tudelft.nl/Other/vacancies.html
Closing date for application: May 1, 2006