High-order essentially nonoscillatory methods based on radial basis functions

Essentially nonoscillatory (ENO) and weighted ENO (WENO) methods on equidistant Cartesian grids are widely employed to solve partial differential equations with discontinuous solutions. However, stable ENO/WENO methods on unstructured grids are less well studied. We propose high-order essentially nonoscillatory methods based on radial basis functions (RBFs) to solve hyperbolic conservation laws. We derive a smoothness indicator that guarantees the satisfaction of the sign property of the resulting interpolant on general one-dimensional grids. Based on this algorithm we introduce an entropy stable arbitrary high-order finite difference method (RBF-TeCNOp) and an entropy stable second order finite volume method (RBF-EFV2) for one-dimensional problems. Hence, we show that methods based on radial basis functions are as powerful as methods based on polynomial reconstruction. Next, we propose a high-order ENO method based on radial basis functions to solve hyperbolic conservation laws on general two-dimensional unstructured grids. The radial basis function reconstruction offers a flexible framework to deal with ill-conditioned cell arrangements. We define a smoothness indicator based the one-dimensional version and a stencil selection algorithm suitable for general meshes. Furthermore, we develop a stable method to evaluate the RBF reconstruction in the finite volume setting to circumvent the stagnation of the error and keep the condition number of the reconstruction bounded. To reduce the computational complexity, we develop the RBF-CWENO method. This method exhibits high-order convergence and robustness when solving challenging problems and is considerably faster. However, the resolution close to shocks and turbulent structures is lower than for the RBF-ENO method. Finally, we present a hybrid high-resolution RBF-ENO method which is based on the RBF-ENO method for unstructured patches and the standard WENO method on structured ones. Furthermore, we introduce a positivity preserving limiter for non-polynomial reconstruction methods that stabilizes the hybrid RBF-ENO method for problems with low density or pressure. We show its robustness on the scramjet inflow problem and a conical aerospike nozzle jet simulation.

Object-oriented finite element programming

New technologies in computer science applied to numerical computations open the door to alternative approaches to mechanical problems using the finite element method. In classical approaches, theoretical developments often become cumbersome and the computer model which follows shows resemblance with the initial problem statement. The first step in the development consists usually in the analysis of the physics of the problem to simulate. The problem is generally described by a set of equations including partial differential equations. This first model is then replaced by successive equivalent or approximated models. The final result consists in a mathematical description of elemental matrices and algorithms describing the matrix form of the problem. The traditional approach consists then in constructing a computer model, generally complex and often quite different from the original mathematical description, thus making further corrections difficult. Therefore, the crucial problem of both the software architecture and the choice of the appropriate programming language is raised. Partially breaking with this approach, we propose a new approach to develop and program finite element formulations. The approach is based on a hybrid symbolic/numerical approach on the one hand, and on a high level software tool, object-oriented programming (supported here by the languages Smalltalk and C++) on the other hand. The aim of this work is to develop an appropriate environment for the algebraic manipulations needed for a finite element formulation applied to an initial boundary value problem, and also to perform efficient numerical computations. The new environment should make it possible to manage al1 the concepts necessary to solve a physical problem: manipulation of partial differential equations, variational formulations, integration by parts, weak forms, finite element approximations… The concepts manipulated therefore remain closely related to the original mathematical framework. The result of these symbolic manipulations is a set of elemental data (mass matrix, stiffness matrix, tangent stiffness matrix,…) to be introduced in a classical numerical code. The object-oriented paradigm is essential to the success of the implementation. In the context of the finite element codes, the object-oriented approach has already proved its capacity to represent and handle complex structures and phenomena. This is confirmed here with the symbolic environment for derivation of finite element formulations in which objects such as expression, integral and variational formulation appear. The link between both the numerical world and the symbolic world is based on an object-oriented concept for automatic programmation of matrix forms derived from the finite element method. As a result, a global environment in which the numerical is capable of evolving, using a language close to the natural mathematical one, is achieved. The potential of the approach is further demonstrated, on the one hand, by the wide range of problems solved in linear mechanics (electrodynamics in 1 and 2D, heat diffusion,…) as well as in nonlinear mechanics (advection dominated 1D problem, Navier Stokes problem), and, on the other hand by the diversity of the formulations manipulated (Galerkin formulations, space-time Galerkin formulations continuous in space and discontinuous in time, generalized Galerkin least-squares formulations).

EPFL1997

Suboptimal colorings and solution of large chromatic scheduling problems

Ivo Blöchliger

Graph Coloring is a very active field of research in graph theory as well as in the domain of the design of efficient heuristics to solve problems which, due to their computational complexity, cannot be solved exactly (no guarantee that an optimal solution will be reported), see [Cul] for a list of over 450 references. The graph coloring problem involves coloring the vertices of a given graph in such a way that two adjacent vertices never share the same color. The goal is to find the smallest number of colors needed to color all vertices in a fashion that satisfies this requirement. This number is called chromatic number and is denoted by Χ. In the first chapter, we present our research on suboptimal colorings and graphs which can be colored in such a way that the number of different colors appearing on the closed neighborhood (a vertex plus its neighbors) of any vertex v is less than Χ. We call such graphs oligomatic. The most interesting result is the following: given a graph and a coloring using Χ + p colors, there exists a vertex v such that there are at least Χ different colors among all vertices which are at a distance of [p/2] + 1 or less from v. We also study the existence of oligomatic graphs in special classes. Additionally, we present results of research on universal graphs which are "generic" oligomatic graphs in the sense that most properties of oligomatic graphs can be analyzed by restricting ourselves to universal graphs. Chapters Two to Four deal with the development of heuristics for two types of graph coloring problems. The tabu search heuristic was a central focus of our research. A tabu search iteratively modifies a candidate solution (which becomes the new candidate solution) with the goal of improving it. In such a procedure it is forbidden (tabu) to undo a modification for a certain number iterations. This mechanism allows to escape from local minima. In Chapter Two, we propose general improvements for tabu search based on some new and simple mechanisms (called reactive tabu schemes) to adapt the tabu tenure (which corresponds to the number of iterations a modification stays tabu). We also introduce distance and similarity measures for graph coloring problems which are needed in iterative procedures such as those of the tabu search. In the third chapter, we present the PARTIALCOL heuristic for the graph coloring problem. This method obtains excellent results compared to other similar methods and is able to color the well known DIMACS [JT96] benchmark graph flat300_28_0 optimally with 28 colors. (The best colorings found to date by other researchers use at least 31 colors.) PARTIALCOL uses partial solutions in its search space, which is a little explored way of approaching the graph coloring problem. Most approaches either use colorings and try to minimize the number of colors, or they use improper colorings (having conflicts, i.e. adjacent vertices which may have the same color) and try to minimize the number of conflicts. In the final chapter, we present a weighted version of the graph coloring problem which has applications in batch scheduling and telecommunication problems. We present two different adaptations of tabu search to the weighted graph coloring problem and test several of the reactive tabu schemes presented in Chapter Two. Further, we devise an adaptive memory algorithm AMA inspired by genetic algorithms. A large set of benchmark graphs with different properties is presented. All benchmark graphs with known optima have been solved to optimality by AMA. A key element of this algorithm is its capacity to determine the number k of colors to be used. Most other graph coloring heuristics need this parameter to be supplied by the user. Considering the results obtained on various graphs, we are confident that the methods developed are very efficient.

EPFL2005

Parallel Algorithms for the Solution of Large-Scale Fluid-Structure Interaction Problems in Hemodynamics

Davide Forti

This thesis addresses the development and implementation of efficient and parallel algorithms for the numerical simulation of Fluid-Structure Interaction (FSI) problems in hemodynamics. Indeed, hemodynamic conditions in large arteries are significantly affected by the interaction of the pulsatile blood flow with the arterial wall. The simulation of fluid-structure interaction problems requires the approximation of a coupled system of Partial Differential Equations (PDEs) and the set up of efficient numerical solution strategies. Blood is modeled as an incompressible Newtonian fluid whose dynamics is governed by the Navier-Stokes equations. Different constituive models are used to describe the mechanical response of the arterial wall; specifically, we rely on hyperelastic isotropic and anistotropic material laws. The finite element method is used for the space discretization of both the fluid and structure problems. In particular, for the Navier-Stokes equations we consider a semi-discrete formulation based on the Variational Multiscale (VMS) method. Among a wide range of possible solution strategies for the FSI problem, here we focus on strongly coupled monolithic approaches wherein the nonlinearities are treated in a fully implicit mode. To cope with the high computational complexity of the three dimensional FSI problem, a parallel solution framework is often mandatory. To this end, we develop a new block parallel preconditioner for the coupled linearized FSI system obtained after space and time discretization. The proposed preconditioner, named FaCSI, exploits the factorized form of the FSI Jacobian matrix, the use of static condensation to formally eliminate the interface degrees of freedom of the fluid equations, and the use of a SIMPLE preconditioner for unsteady Navier-Stokes equations. In FSI problems, the different resolution requirements in the fluid and structure physical domains, as well as the presence of complex interface geometries make the use of matching fluid and structure meshes problematic. In such situations, it is much simpler to deal with discretizations that are nonconforming at the interface, provided however that the matching conditions at the interface are properly fulfilled. In this thesis we develop a novel interpolation-based method, named INTERNODES, for numerically solving partial differential equations by Galerkin methods on computational domains that are split into two (or several) subdomains featuring nonconforming interfaces. By this we mean that either a priori independent grids and/or local polynomial degrees are used to discretize each subdomain. INTERNODES can be regarded as an alternative to the mortar element method: it combines the accuracy of the latter with the easiness of implementation in a numerical code. The aforementioned techniques have been applied for the numerical simulation of large-scale fluid-structure interaction problems in the context of biomechanics. The parallel algorithms developed showed scalability up to thousands of cores utilized on high performance computing machines.

EPFL2016