National Technical University of Athens (NTUA)
Αδρομερής προσομοίωση περίπλοκων φυσικοχημικών και βιολογικών συστημάτων με μεθόδους ανάλυσης πολλαπλών κλιμάκων, διακλάδωσης και ομοιοτυπίας
Abstract
dc:descriptionOne of the main challenges in contemporary scientific problems is the simulation of complicated systems, the behavior of which is determined by the interaction of multiple time and space scales. Especially, in engineering applications, the main challenge is to predict the macroscopic (coarse) behavior of physicochemical and biological systems starting from models which describe processes at the microscopic level. The main target of the present dissertation is to complete the deficient synergy between traditional numerical methods, applied at macroscopic level, and microscopic simulations. A computational framework is developed and permits the efficient study of such systems applying multiscale methods, bifurcation analysis and self-similar solution computations. The generality and adaptability of these methods enables their application in a large variety of problems. We studied cell population balance models, where the main interest is to investigate the cell heterogeneity effect on the phenotype of cell populations, bearing the same gene regulatory network. This study sets of a way to bridge the gap between scales of different magnitude (cell vs cell population). The application of bifurcation analysis showed the existence of a range of parameter values (such as the extracellular inducer concentration) where the cell population phenotype exhibits bi-stability. The bi-stability range appears also at the single cell level, however it is restricted due to the cell heterogeneity effect. We analyzed the instability mechanisms appearing in polymer extrusion processes where, at a certain range of volumetric flow rate, oscillating solutions are observed. This behavior has also been observed experimentally; the application of bifurcation analysis indicated a range of volumetric flow rate values, where stable periodic and steady state solutions co-exist. In this dissertation, we mainly focus on problems with continuous symmetries and especially problems with translational invariance and self-similarity. The main feature of self-similar solutions, is that they evolve towards scales of larger magnitude (blow up solutions) or scales of smaller magnitude (decaying solutions) preserving their shape invariant under certain transformations. The computation of self-similar solutions accommodates the efficient study of such problems, avoiding the transient simulations at critical times where the scale of the solution differs by orders of magnitude from the initial condition. We develop and apply the template-based method which permits the construction of a co-evolving coordinate framework, where the rescaled solution evolves towards the invariant shape, namely self-similar or translationally invariant, solution. In problems with translational invariance, we examine the Nagumo equation, where the traveling invariant solution and the moving front velocity are computed. The method can also be adapted and modified for the computation of self-similar solutions and their similarity exponents; cases examined are one-dimensional and two-dimensional diffusion equations. It can be also adapted for problems with combined translational and scale invariance (Burgers equation). The modified projective integration algorithm is developed for the acceleration of temporal computations in problems with continuous symmetries. The co-evolving frame is determined through the application of the template-based method and the projection is applied in this frame. The efficiency of projective integration in the co-evolving frame is demonstrated for the Nagumo equation and the one dimensional diffusion equation. The “equation-free” method is applied in systems simulated at micro-scale level. This technique aspires to bridge the gap between the observed time and spatial scales, extracting coarse information from the execution of appropriately initialized computational experiments. The key idea in this computational framework is to construct the so-called coarse time integrator, where traditional computational methods used for the study of partial differential equation problems can also be applied. We applied the template-based method for the computation of coarse self-similar solutions in a particle diffusion problem, the dynamic behaviour of which is determined by Monte Carlo algorithms. The application of the modified coarse projective integration accelerates the temporal computations and preserves the accuracy of the obtained results. The modified coarse projective integration scheme is applied in a diffusion-reaction problem with moving front solutions. The dynamic simulation is performed by a kinetic Monte Carlo algorithm. We use Fourier coefficients for the coarse description of the system and a co-evolving coordinate framework is developed using trigonometric template functions.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2007
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kavousanakis, Michail
- Καβουσανάκης, Μιχαήλ
Subjects
dc:subject × 12Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/16288