National Technical University of Athens (NTUA)
Numerical analysis in parameter identification problems
Abstract
dc:descriptionThis thesis is devoted to the derivation of rigorous residual-type a posteriori error estimates for the Robin coefficient parameter identification bilinear inverse problem subject to elliptic and parabolic partial differential equations as constraints. Motivated by the optimization problems at have, we first prove L2-like norm residual typea posteriori error estimates for finite element methods for elliptic partial differential equations with non-essential boundary conditions, such as Neumann or Robin type. To ensure the proof of lower bounds (efficiency), a non-standard mesh-dependent L2-like norm is used for the error. The proof of lower bounds requires a carefully constructed C1-conforming’bubble’-function. A series of extensive numerical experiments is presented, showcasing the good performance of the estimators. The main objective is the proof of a posteriori error estimates for the stationary identification inverse problem of the Robin coefficient. More specifically, with the use of the second order sufficient condition, initially we prove estimates with upper bounds in L2-norm and later on, we derive a a posteriori error estimates with upper and lower bounds in H1-norm,in a fully discrete scheme. We also prove the corresponding a posteriori error estimates in the context of the variational approach and for a, so-called, semi discrete scheme, resulting in upper and lower bounds in the H1-norm.Finally, we prove an a posteriori error estimate for the inverse problem of estimating the spatially-and-temporally dependent heat transfer coefficient. We work on a fully-discrete scheme based on the discontinuous-in time-Galerkin dG(0) framework and finite elements in space and we carefully design elliptic and time reconstructions for the discrete state and ad joint equation, to obtain the desirable result. The main challenge in the proof of the estimators is, again, the bilinear nature of the problem, which multiplying the control (Robin coefficient) by the state variable. To circumvent this, a second order sufficient optimality condition is postulated, offering the required stability to facilitate the proof.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Papadopoulos, Vasileios
- Παπαδόπουλος, Βασίλειος
Subjects
dc:subject × 24- Αριθμητική ανάλυση
- Βέλτιστος έλεγχος
- Μερικές διαφορικές εξισώσεις
- Εκ των υστέρων εκτιμήσεις σφάλματος
- Αντίστροφα προβλήματα
- Ταυτοποίηση παραμέτρων
- Συνοριακές συνθήκες Robin
- Μη κυρτή βελτιστοποίηση
- Numerical analysis
- Οptimal control
- Partial differential equations
- A posteriori error analysis
- Inverse problems
- Parameter identification
- Robin boundary conditions
- Non convex optimization
- Μαθηματικά
- Φυσικές Επιστήμες
- Υπολογιστικά μαθηματικά
- Έλεγχος και Βελτιστοποίηση
- Mathematics
- Natural Sciences
- Computational Mathematics
- Control and Optimization
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/57177