One important aspect of our present age of monolithic high speed computers is the computer's capability to solve complex problems hitherto impossible to tackle due to their complexity. This paper explains how to use a. digital computer to solve a specific type of problem; specifically, to find the inverse solution of a in the elliptical equation V2U(x,y) = g(a,U,x,y), with appropriate boundary conditions. This equation is very useful in the electronics field. The knowns are the complete set of boundary values of U(x,y) and a set of observations taken on internal points of U(x,y). Given this information, plus the specific form of the governing equation, we can solve for the unknown a. Once the computer program has been written using the technique of quasilinearization, Newton’S convergence method, discrete invariant imbedding, and the use of sensitivity functions, then we take data from the computer results and analyse it for proper convergence. This data shows that there are definite limits to the usefulness and capability of the technique. One of the results of this study is the observation that it is important to the proper functioning of this problem solving technique that the observations taken on U(x,y) are placed in the most efficient locations with the most efficient geometry in the region of largest effectiveness. Another result deals with the number of observation points used: too few gives insufficient information for proper program functioning, and too many tends to saturate the effectiveness of the observations. Thus this paper has two objectives. first to develop the technique and secondly to analyse the results from the realization of the technique through the use of a computer.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Civil Engineering, 1981. / MICROFICHE COPY AVAILABLE IN ARCHIVES AND ENGINEERING. / Bibliography: leaves 222-223. / by John Lambros Tassoulas. / Ph.D.

Yes / Seeking efficient solutions to nonlinear boundary value problems is a crucial challenge in the mathematical modelling of many physical phenomena. A well-known example of this is solving the Biharmonic equation relating to numerous problems in fluid and solid mechanics. One must note that, in general, it is challenging to solve such boundary value problems due to the higher-order partial derivatives in the differential operators. An artificial neural network is thought to be an intelligent system that learns by example. Therefore, a well-posed mathematical problem can be solved using such a system. This paper describes a mesh free method based on a suitably crafted deep neural network architecture to solve a class of well-posed nonlinear boundary value problems. We show how a suitable deep neural network architecture can be constructed and trained to satisfy the associated differential operators and the boundary conditions of the nonlinear problem. To show the accuracy of our method, we have tested the solutions arising from our method against known solutions of selected boundary value problems, e.g., comparison of the solution of Biharmonic equation arising from our convolutional neural network subject to the chosen boundary conditions with the corresponding analytical/numerical solutions. Furthermore, we demonstrate the accuracy, efficiency, and applicability of our method by solving the well known thin plate problem and the Navier-Stokes equation.

We study an inverse problem which seeks to image the internal conductivity map of a body by one measurement of boundary and interior data. In our study the interior data is the magnitude of the current density induced by electrodes. Access to interior measurements has been made possible since the work of M. Joy et al. in early 1990s and couples two physical principles: electromagnetics and magnetic resonance. In 2007 Nachman et al. has shown that it is possible to recover the conductivity from the magnitude of one current density field inside. The method now known as Current Density Impedance Imaging is based on solving boundary value problems for the 1-Laplacian in an appropriate Riemann metric space. We consider two types of methods: the ones based on level sets and a variational approach, which aim to solve specific boundary value problem associated with the 1-Laplacian. We will address the Cauchy and Dirichlet problems with full and partial data, and also the Complete Electrode Model (CEM). The latter model is known to describe most accurately the voltage potential distribution in a conductive body, while taking into account the transition of current from the electrode to the body. For the CEM the problem is non-unique. We characterize the non-uniqueness, and explain which additional measurements fix the solution. Multiple numerical schemes for each of the methods are implemented to demonstrate the computational feasibility.

A modified source and dipole panel method to calculate the flow properties around an oscillating arbitrary body in the presence of a free surface is proposed. To demonstrate the feasibility of the method the problem of a pulsating line source submerged under a free surface is treated. The technique chosen is based on Green's identity whereby the boundary-value problem is expressed as a boundary integral equation which is solved numerically. The near field of the water surface is represented by singularity panels with constant strength. The work was motivated by the reported large computing times for existing programs using Green's functions satisfying the free surface boundary condition. The present approach uses free-space Green's function. The free surface boundary condition is applied to surface singularity panels using Green's theorem. Thus free surface effects are included in the solution while panel integrations are simplified considerably by the use of simpler Green's function. The matrix equations resulting from Green's identity were solved by using IMSL routines for Gaussian Elimination, and the behavior of the influence coefficient matrix was tested by using LINPACK routines. The depth of the submerged-source and wave number was kept constant while the length of near field and the number of panels per wavelength was varied systematically. A minimum of 10 panels per wavelength and paneled water surface length of 2 wavelengths gives good agreement with the known exact solution. Computing times were low, indicating the feasibility of the technique for application to unsteady ship problems. / Ph. D.

Dans cette thèse, nous étudions des problèmes aux limites pour les systèmes elliptiques sous forme divergence avec coefficients complexes dans L^{infty}. Nous prouvons des estimations a priori, discutons de la solvabilité et d'extrapolation de la solvabilité. Nous utilisons une transformation via des équations Cauchy-Riemann généralisées due à P. Auscher, A. Axelsson et A. McIntosh. On peut résoudre les équations Cauchy-Riemann généralisées via la semi-groupe engendré par un opérateur différentiel perturbé d'ordre un de type Dirac. A l'aide du semi-groupe, nous étudions la théorie L^{p} avec une discussion sur la bisectorialité, le calcul fonctionnel holomorphe et les estimations hors-diagonales pour des opérateurs dans le calcul fonctionnel. En particulier, nous développons une théorie L^{p}-L^{q} pour des opérateurs dans le calcul fonctionnel d'opérateur de type Dirac perturbé. Les problèmes de Neumann, Régularité et Dirichlet se formulent avec des estimations quadratiques et des estimations pour la fonction maximale nontangentielle. Cela conduit à à démontrer de telles estimations pour le semi-groupe d'opérateur de Dirac Pour cela, nous utilisons les espaces Hardy associés et les identifions dans certains cas avec des sous-espaces des espaces de Hardy et Lebesgue classiques. Nous obtenons enfin des estimations a priori pour les problème aux limites via une extension utilisant des espaces de Sobolev associés. Nous utilisons les estimations a priori pour une discussion sur la solvabilité des problèmes aux limites et montrer un théorème d'extrapolation de la solvabilité. / In this this thesis we study boundary value problems for elliptic systems in divergence form with complex coefficients in L^{\infty}. We prove a priori estimates, discuss solvability and extrapolation of solvability. We use a transformation to generalized Cauchy-Riemann equations due to P. Auscher, A. Axelsson, and A. McIntosh. The generalized Cauchy-Riemann equations can be solved by the semi-group generated by a perturbed first order Dirac/differential operator. In relation to semi-group theory we setup the L^p theory by a discussion of bisectoriality, holomorphic functional calculus and off-diagonal estimates for operators in the functional calculus. In particular, we develop an L^p-L^q theory for operators in the functional calculus of the first order perturbed Dirac/differential operators. The formulation of Neumann, Regularity and Dirichlet problems involve square function estimates and nontangential maximal function estimates. This leads us to discuss square function estimates and nontangential maximal function estimates involving operators in the functional calculus of the perturbed first order Dirac/differential operator. We discuss the related Hardy spaces associated to operators and prove identifications by subspaces of classical Hardy and Lebesgue spaces. We obtain the a priori estimates by an extension of the square function estimates and nontangential maximal function estimates to Sobolev spaces associated to operators. We use the a priori estimates for a discussion of solvability and extrapolation of solvability.

For the numerical simulation of many problems of engineering interest, it is desirable to have an automated mesh adaption tool. This is important especially for problems characterized by anisotropic features and require mesh clustering in the direction of high gradients. Another significant issue in meshing emerges in unsteady simulations with moving boundaries, where the boundary motion has to be accommodated by deforming the computational grid. Similarly, there exist problems where current mesh needs to be adapted to get more accurate solutions. To solve these problems, we propose three novel procedures. In the first part of this work, we present an optimization procedure for three-dimensional anisotropic tetrahedral grids based on metric-driven h-adaptation. Through the use of topological and geometrical operators, the mesh is iteratively adapted until the final mesh minimizes a given objective function. We propose an optimization process based on an ad-hoc application of the simulated annealing technique, which improves the likelihood of removing poor elements from the grid. Moreover, a local implementation of the simulated annealing is proposed to reduce the computational cost. Many challenging unsteady multi-physics problems are characterized by moving boundaries and/or interfaces. When the boundary displacements are large, degenerate elements are easily formed in the grid such that frequent remeshing is required. We propose a new r-adaptation technique that is valid for all types of elements (e.g., triangle, tet, quad, hex, hybrid) and deforms grids that undergo large imposed displacements at their boundaries. A grid is deformed using a network of linear springs composed of edge springs and a set of virtual springs. The virtual springs are constructed in such a way as to oppose element collapsing. Both frequent remeshing, and exact-pinpointing of clustering locations are great challenges of numerical simulations, which can be overcome by adaptive meshing algorithms. Therefore, we conclude this work by defining a novel mesh adaptation technique where the entire mesh is adapted upon application of a force field in order to comply with the target mesh or to get more accurate solutions. The method has been tested for two-dimensional problems of a-priori metric definitions as well as for oblique shock clusterings.

M. Tech. : Mathematical Technology. / Attempts will be made to extend the elementary quarter-mass models (for instance Gillepse, 1992, [5]; Kiecke & Nielsen, 2000, [6] and Singiresu, 2004, [7]) of a motor vehicle suspension system to include the radial vibrations of a rubber tyre in the model. Tangential vibrations of the tyre surface were investigated by Bekker (2009, [8]) and the possible incorporation of such vibrations into a suspension model invites the possibility of future study.

In several mathematical models of physical or technical processes there are non-local boundary-value problems in terms of partial differential equations with integral conditions. In this article we consider hyperbolic differential equations of second order in the rectangle with some integral conditions and their relationship to boundary-value problems for some certain type of loaded equations. / Dieser Beitrag ist mit Zustimmung des Rechteinhabers aufgrund einer (DFG-geförderten) Allianz- bzw. Nationallizenz frei zugänglich.

No description available.