In this paper an on-line algorithm for the Rectangle Packing
Problem is presented. The method is designed to be able to accept
or reject incoming boxes to maximize efficiency. We provide a wide
computational analysis showing the behavior of the proposed
algorithm as well as a comparison with existing off-line
heuristics.
In this work we study a finite dynamical system for the description of the
bifurcation pattern of the convection flow of a fluid between two parallel
horizontal planes which undergoes a {\em horizontal} gradient of temperature
({\em horizontal} convection flow). Although in the two-dimensional case
developed
here,literature reports as well a long list of analytical and numerical
solutions to this problem, the peculiar aim of this work makes it worthwhile.
Actually we develop the route that Saltzman (1962) \cite{Sal62} and
Lorenz (1963) \cite{Lor63} proposed for the {\em vertical}
convection flow that started successfully the approach to finite dynamical
systems. We obtain steady-to-steady and steady-to-periodic bifurcations in
qualitative agreement with already published results. At first we adopt the
non-dimensional scheme used by Saltzman and Lorenz; as it admits also
physically meaningless solutions, we introduce a different set of
reference quantities so overcoming this drawback.
The existence of global smooth solutions to the multi-dimensional
hydrodynamic model for plasmas of electrons and positively charged ions
is shown under the assumption that the initial densities are close to
a constant. The model consists of the conservation laws for the particle densities and the current densities, coupled to the Poisson equation for the electrostatic potential.
Furthermore, it is proved that the particle densities converge
exponentially fast to the (constant) steady state. The proof uses a higher-order energy method inspired from extended thermodynamics.
Hydrodynamic model
global existence
exponential stabilit
plasmas
We derive an asymptotic equation that describes the propagation
of weakly nonlinear surface waves on a tangential discontinuity
in incompressible magnetohydrodynamics. The equation is similar
to, but simpler than, previously derived asymptotic
equations for weakly nonlinear Rayleigh waves in elasticity.
The most interesting feature of the surface waves is that their
nonlinear self-interaction is nonlocal.
In refined network analysis, a compact network model is
combined with distributed models for semiconductor devices in a
multiphysics approach. For linear RLC networks containing diodes as
distributed devices, we construct a mathematical model that combines the
differential-algebraic network equations of the circuit with elliptic
boundary value problems modeling the diodes.
For this mixed initial-boundary value problem of
partial differential-algebraic equations a first existence result is given.
The global existence of smooth solutions of the Cauchy
problem for the $N$-dimensional Euler-Poisson model for
semiconductors is established,
under the assumption
that the initial data is a perturbation of a stationary solution
of the drift-diffusion equations with zero electron velocity,
which is proved to be unique.
The resulting evolutionary solutions converge asymptotically in time to
the unperturbed state.
The singular relaxation limit is also discussed.
We perform a multiple time scale, single space scale analysis of a compressible
fluid in a time-dependent domain, when the time variations of the boundary
are small with respect to the acoustic velocity.
We introduce an average operator with respect to the fast time.
The averaged leading order variables
satisfy modified incompressible equations, which are
coupled to linear acoustic equations with respect to the fast time.
We discuss possible initial-boundary data for the asymptotic equations
inherited from the initial-boundary data for the compressible equations.
We consider the Cauchy problem for a general one dimensional $n\times n$ hyperbolic
symmetrizable system of balance laws. It is well known that, in many physical
examples, for instance for the isentropic Euler system with damping, the
dissipation due to the source term may prevent the shock formation, at least
for smooth and small initial data.
Our main goal is to find a set of general
and realistic sufficient conditions to guarantee the global existence of
smooth solutions, and possibly to investigate their asymptotic behavior.
For systems which are entropy dissipative, a quite natural generalization of
the Kawashima condition for hyperbolic-parabolic systems can be given. In
this paper, we first propose a general framework to set this kind of problems, by using the so-called entropy variables. Therefore, we pass to prove some
general statements about the global existence of smooth solutions, under
different sets of conditions. In particular, the present approach is suitable for dealing with most of the physical examples of systems with a relaxation extension. Our main tools will be some refined energy
estimates and the use of a suitable version of the
Kawashima condition.
Sistemi iperbolici
entropie dissipative
stabilita' globale
sistemi di rilassame
Entropy solutions to a strongly degenerate anisotropic convection-diffusion equation, with application to the backward-forward stochastic differential utility
We study the deterministic counterpart of a backward-forward stochastic differential utility, which has recently been characterized as the solution to the Cauchy problem related to a PDE of degenerate parabolic type in two spatial variables, with a rank-1 diffusion and a conservative first order term. We first establish a local existence result for strong solutions and a continuation principle, and we produce a counterexample showing that, in general, strong solutions fail to be globally.
We introduce a differential model to study damage accumulation processes in
presence of chemical reactions. The influence of micro-structure
leads to a nonlinear parabolic system characterized by the presence of a characteristic length. Here, we first present an analytical description of the qualitative behavior of solutions
which blow-up in finite time. Numerical simulations are given to describe the shape
of solutions near the rupture time and the influence of the chemical reagents.
Like in the non reactive model, the failure of the material occurs in a region of positive measure, due to the diffusive effects of the micro-structure,
although some localization phenomena are observed.
Moreover, if we increase the chemical concentration beyond a given threshold, which depends on the specific conditions of the material, we observe a strong acceleration in the damage process.
This paper deals with the development of a scientific computing environment for differential field simulation. We mean a modelling and simulation environment based on partial differential equations and their numerical solution as powerful and widely used technique for mathematical and computational investigation of application problems. We have been developing grid generation algorithms, numerical solvers of PDE systems, along with advanced visualization techniques, to numerically compute and evaluate field variables by exploiting user-friendly interaction. In this paper, we model the complete cycle of the visual computational simulation as reference framework and we illustrate advances in the environment development. We describe a few computational components by focusing on two fundamental substeps often conscurring to simulation processes, the image segmentation and grid generation. We introduce differential equation systems, developed combination of computational methods and recent algorithmic advances. A few application results are detailed, and shown by figures, for segmentation test problems.
The detailed analysis of a 1D-model for fluid flows in porous media with piecewise constant permeability clearly shows that variable permeability may lead to the phenomenon of oil trapping.
We consider the problem of writing Glimm type interaction estimates for the
hyperbolic system
\begin{equation}\label{E:abs0}
u_t + A(u) u_x = 0.
\end{equation}
%only assuming that $A(u)$ is strictly hyperbolic.
The aim of these estimates is to prove that there is Glimm-type
functional $Q(u)$ such that
\begin{equation}\label{E:abs1}
\TV(u) + C_1 Q(u) \ \text{is lower semicontinuous w.r.t.} \ L^1-\text{norm},
\end{equation}
with $C_1$ sufficiently large, and $u$ with small BV norm.
In the first part we analyze the more general case of quasilinear hyperbolic
systems. We show that in general this result is not true if the system
is not in conservation form: there are Riemann solvers, identified by selecting
an entropic conditions on the jumps, which do not
satisfy the Glimm interaction estimate \eqref{E:abs1}.
Next we consider hyperbolic systems in conservation form, i.e. $A(u) = Df(u)$.
In this case, there is only one entropic Riemann solver, and we
prove that this particular
Riemann solver satisfies \eqref{E:abs1} for a particular functional
$Q$, which we construct explicitly. The main novelty here is that we suppose
only the Jacobian matrix $Df(u)$ strictly
hyperbolic, without any assumption on the number of inflection points of $f$.
These results are achieved by an analysis of the growth of $\TV(u)$ when
nonlinear waves of \eqref{E:abs0} interact, and the
introduction of a Glimm type functional $Q$, similar but not equivalent to Liu's
interaction functional \cite{liu:admis}.
As strictly hyperbolic system of conservation laws of the form $$ u_{t}+f(u)_x =0 , \quad u(0,x)=\bar u (x)$$ is considered, where $ u \in\bbfR^N$, $f:\bbfR^N \rightarrow\bbfR^N$ is smooth, especially from a numerical point of view, that means, a semidiscrete upwind scheme of this equation is investigated. If we suppose that the initial data $\bar u (x) $ of this problem have small total variation the author proves that the solution of the upwind scheme $$ {\partial u(t,x) \over \partial t} + { ( f(u(t,x))-f(u(t,x-\varepsilon))) \over \varepsilon} =0 $$ has uniformly bounded variation (BV) norm independent on $t$ and $\varepsilon$. Moreover the Lipschitz-continuous dependence of the solution of the upwind scheme $u^{\varepsilon}(t)$ on the initial data is proved. This solution $u^{\varepsilon}(t)$ converges in $ L_1$ to a weak solution of the corresponding hyperbolic system as $ \varepsilon \rightarrow 0$. This weak solution coincides with the trajectory of a Riemann semigroup which is uniquely determined by the extension of Liu's Riemann solver to general hyperbolic systems.
The paper concerns with a hyperbolic system of conservation laws in one space variable $$ u_t + f(u)_x = 0,\qquad u(0,x) = u_0(x), $$ where $ u \in \Bbb R^n$, $f:\Omega \subseteq \Bbb R^n \rightarrow \Bbb R^n.$ Let $ K_0 \subset \Omega $ be a compact and let $\delta_1 > 0 $ be sufficiently small such that $K_1 = \{ u \in \Bbb R^n: \text{dist}(u,K_0) \leq \delta_1\}\subset \Omega.$ \par Assuming that the Jacobian matrix $A = Df$ is uniformly strictly hyperbolic in $K_1, u_0(-\infty) \in K_0$ and that the total variation of $u_0$ is sufficiently small, then there exists a unique ``entropic" solution $u: [0,+\infty) \rightarrow BV(\Bbb R,\Bbb R^n).$
We derive and study Well-Balanced schemes for quasimonotone discrete kinetic models. By means of a rigorous localization procedure, we reformulate the collision terms as nonconservative products and solve the resulting Riemann problem whose solution is self-similar. The construction of an Asymptotic Preserving (AP) Godunov scheme is straightforward and various compactness properties are established within different scalings. At last, some computational results are supplied to show that this approach is realizable and efficient on concrete $2 \times 2$ models.
kinetic model
discrete velocities
Well-Balanced scheme
diffusive regime
Barenblatt solution
Two systems of hyperbolic equations, arising in the multiphase
semiclassical limit of the linear Schr\"odinger equations, are
investigated. One stems from a Wigner measure analysis and uses a
closure by the Delta functions, whereas the other relies on the
classical WKB expansion and uses the Heaviside functions for closure.
The two resulting moment systems are weakly
and non-strictly hyperbolic respectively. They provide two
different Eulerian methods able to reproduce superimposed signals with a
finite number of phases. Analytical properties of these moment
systems are investigated and compared. Efficient numerical
discretizations and test-cases with increasing difficulty are
presented.