Thu, 25.05.2023 15:30

RICAM Colloquium - Observer based data assimilation for one dimensional, barotropic Euler equations on networks

RICAM Colloquium - Computational Methods for PDEs – Prof. Jan Giesselmann

Date: Thursday, May 25, 15:30

Place: SP 416-2


In this talk, we discuss state estimation for systems described by one dimensional, barotropic Euler equations on networks. These are a frequently used as models for gas flows in pipeline networks. We are interested in predicting the system state based on a combination of models and measurements that have been taken over a long period of time.

We will consider two scenarios where the system state can only be measured partially: In the first scenario, we assume that measurements can be made only at nodes of the network. In the second scenario, we assume that measurements can be made everywhere but just one field, e.g. velocity, can be measured. In both scenarios we construct so called observers, i.e. models that combine a PDE model that encodes our understanding of the underlying physics with measurement data that are included as boundary data or source terms.

Our goal is to show that the solution of the observer system converges towards the state of the original system exponentially for long times. We achieve this     under certain assumptions on the size and regularity of the solution to the original system and restrictions on the network topology. On a technical level, in the first scenario the proof relies on Riemann invariants and the their evolution along characteristics whereas in the second scenario we use estimates of an extended relative energy.

This is joint work with Martin Gugat (Erlangen) and Teresa Kunkel (Darmstadt).