Showing papers on "Discretization published in 1995"

Statistical Mechanics of Dissipative Particle Dynamics.

Abstract: The stochastic differential equations corresponding to the updating algorithm of Dissipative Particle Dynamics (DPD), and the corresponding Fokker-Planck equation are derived. It is shown that a slight modification to the algorithm is required before the Gibbs distribution is recovered as the stationary solution to the Fokker-Planck equation. The temperature of the system is then directly related to the noise amplitude by means of a fluctuation-dissipation theorem. However, the correspondingly modified, discrete DPD algorithm is only found to obey these predictions if the length of the time step is sufficiently reduced. This indicates the importance of time discretisation in DPD.

Supervised and unsupervised discretization of continuous features

Abstract: Many supervised machine learning algorithms require a discrete feature space. In this paper, we review previous work on continuous feature discretization, identify defining characteristics of the methods, and conduct an empirical evaluation of several methods. We compare binning, an unsupervised discretization method, to entropy-based and purity-based methods, which are supervised algorithms. We found that the performance of the Naive-Bayes algorithm significantly improved when features were discretized using an entropy-based method. In fact, over the 16 tested datasets, the discretized version of Naive-Bayes slightly outperformed C4.5 on average. We also show that in some cases, the performance of the C4.5 induction algorithm significantly improved if features were discretized in advance; in our experiments, the performance never significantly degraded, an interesting phenomenon considering the fact that C4.5 is capable of locally discretizing features.

Chi2: feature selection and discretization of numeric attributes

Abstract: Discretization can turn numeric attributes into discrete ones. Feature selection can eliminate some irrelevant attributes. This paper describes Chi2 a simple and general algorithm that uses the /spl chi//sup 2/ statistic to discretize numeric attributes repeatedly until some inconsistencies are found in the data, and achieves feature selection via discretization. The empirical results demonstrate that Chi/sup 2/ is effective in feature selection and discretization of numeric and ordinal attributes.

The pseudospectral Legendre method for discretizing optimal control problems

Abstract: This paper presents a computational technique for optimal control problems including state and control inequality constraints. The technique is based on spectral collocation methods used in the solution of differential equations. The system dynamics are collocated at Legendre-Gauss-Lobatto points. The derivative x/spl dot/(t) of the state x(t) is approximated by the analytic derivative of the corresponding interpolating polynomial. State and control inequality constraints are collocated at Legendre-Gauss-Lobatto nodes. The integral involved in the definition of the performance index is discretized based on the Gauss-Lobatto quadrature rule. The optimal control problem is thereby converted into a mathematical programming program. Thus existing, well-developed optimization algorithms may be used to solve the transformed problem. The method is easy to implement, capable of handling various types of constraints, and yields very accurate results. Illustrative examples are included to demonstrate the capability of the proposed method, and a comparison is made with existing methods in the literature. >

Finite Element Algorithms for Contact Problems

Abstract: The numerical treatment of contact problems involves the formulation of the geometry, the statement of interface laws, the variational formulation and the development of algorithms. In this paper we give an overview with regard to the different topics which are involved when contact problems have to be simulated. To be most general we will derive a geometrical model for contact which is valid for large deformations. Furthermore interface laws will be discussed for the normal and tangential stress components in the contact area. Different variational formulations can be applied to treat the variational inequalities due to contact. Several of these different techniques will be presented. Furthermore the discretization of a contact problem in time and space is of great importance and has to be chosen with regard to the nature of the contact problem. Thus the standard discretization schemes will be discussed as well as techiques to search for contact in case of large deformations.

DANTSYS: A diffusion accelerated neutral particle transport code system

Abstract: The DANTSYS code package includes the following transport codes: ONEDANT, TWODANT, TWODANT/GQ, TWOHEX, and THREEDANT. The DANTSYS code package is a modular computer program package designed to solve the time-independent, multigroup discrete ordinates form of the boltzmann transport equation in several different geometries. The modular construction of the package separates the input processing, the transport equation solving, and the post processing (or edit) functions into distinct code modules: the Input Module, one or more Solver Modules, and the Edit Module, respectively. The Input and Edit Modules are very general in nature and are common to all the Solver Modules. The ONEDANT Solver Module contains a one-dimensional (slab, cylinder, and sphere), time-independent transport equation solver using the standard diamond-differencing method for space/angle discretization. Also included in the package are solver Modules named TWODANT, TWODANT/GQ, THREEDANT, and TWOHEX. The TWODANT Solver Module solves the time-independent two-dimensional transport equation using the diamond-differencing method for space/angle discretization. The authors have also introduced an adaptive weighted diamond differencing (AWDD) method for the spatial and angular discretization into TWODANT as an option. The TWOHEX Solver Module solves the time-independent two-dimensional transport equation on an equilateral triangle spatial mesh. The THREEDANT Solver Module solves the time independent, three-dimensional transport equation for XYZ and RZ{Theta} symmetries using both diamond differencing with set-to-zero fixup and the AWDD method. The TWODANT/GQ Solver Module solves the 2-D transport equation in XY and RZ symmetries using a spatial mesh of arbitrary quadrilaterals. The spatial differencing method is based upon the diamond differencing method with set-to-zero fixup with changes to accommodate the generalized spatial meshing.

Numerical Integration of the Shallow-Water Equations on a Twisted Icosahedral Grid. Part I: Basic Design and Results of Tests

Abstract: The streamfunction-velocity potential form of shallow-water equations, implemented on a spherical geodesic grid, offers an attractive solution to many of the problems associated with fluid-flow simulations in a spherical geometry. Here construction of a new type of spherical geodesic grid is outlined, and discretization of the equations is explained. The model is subjected to the NCAR suite of seven test cases for shallow-water models.

Analysis of thin plates by the element-free Galerkin method

Abstract: A meshless approach to the analysis of arbitrary Kirchhoff plates by the Element-Free Galerkin (EFG) method is presented. The method is based on moving least squares approximant. The method is meshless, which means that the discretization is independent of the geometric subdivision into “finite elements”. The satisfaction of the C 1 continuity requirements are easily met by EFG since it requires only C 1 weights; therefore, it is not necessary to resort to Mindlin-Reissner theory or to devices such as discrete Kirchhoff theory. The requirements of consistency are met by the use of a quadratic polynomial basis. A subdivision similar to finite elements is used to provide a background mesh for numerical integration. The essential boundary conditions are enforced by Lagrange multipliers. It is shown, that high accuracy can be achieved for arbitrary grid geometries, for clamped and simply-supported edge conditions, and for regular and irregular grids. Numerical studies are presented which show that the optimal support is about 3.9 node spacings, and that high-order quadrature is required.

Adaptive finite element methods for parabolic problems II: optimal error estimates in L ∞ L 2 and L ∞ L ∞

Abstract: Optimal error estimates are derived for a complete discretization of linear parabolic problems using space–time finite elements. The discretization is done first in time using the discontinuous Galerkin method and then in space using the standard Galerkin method. The underlying partitions in time and space need not be quasi uniform and the partition in space may be changed from time step to time step. The error bounds show, in particular, that the error may be controlled globally in time on a given tolerance level by controlling the discretization error on each individual time step on the same (given) level, i.e., without error accumulation effects. The derivation of the estimates is based on the orthogonality of the Galerkin procedure and the use of strong stability estimates. The particular and precise form of these error estimates makes it possible to design efficient adaptive methods with reliable automatic error control for parabolic problems in the norms under consideration.

Class-dependent discretization for inductive learning from continuous and mixed-mode data

Abstract: Inductive learning systems can be effectively used to acquire classification knowledge from examples. Many existing symbolic learning algorithms can be applied in domains with continuous attributes when integrated with a discretization algorithm to transform the continuous attributes into ordered discrete ones. In this paper, a new information theoretic discretization method optimized for supervised learning is proposed and described. This approach seeks to maximize the mutual dependence as measured by the interdependence redundancy between the discrete intervals and the class labels, and can automatically determine the most preferred number of intervals for an inductive learning application. The method has been tested in a number of inductive learning examples to show that the class-dependent discretizer can significantly improve the classification performance of many existing learning algorithms in domains containing numeric attributes. >

Implicit time discretization for the mean curvature flow equation

Abstract: In this paper we apply the method of implicit time discretization to the mean curvature flow equation including outer forces. In the framework ofBV-functions we construct discrete solutions iteratively by minimizing a suitable energy-functional in each time step. Employing geometric and variational arguments we show an energy estimate which assures compactness of the discrete solutions. An additional convergence condition excludes a loss of area in the limit. Thus existence of solutions to the continuous problem can be derived. We append a brief discussion of the related Mullins-Sekerka equation.

On the abstract theory of additive and multiplicative Schwarz algorithms

Abstract: In recent years, it has been shown that many modern iterative algorithms (multigrid schemes, multilevel preconditioners, domain decomposition methods etc.) for solving problems resulting from the discretization of PDEs can be interpreted as additive (Jacobi-like) or multiplicative (Gauss-Seidel-like) subspace correction methods. The key to their analysis is the study of certain metric properties of the underlying splitting of the discretization space \(V\) into a sum of subspaces \(V_j\) and the splitting of the variational problem on \(V\) into auxiliary problems on these subspaces. In this paper, we propose a modification of the abstract convergence theory of the additive and multiplicative Schwarz methods, that makes the relation to traditional iteration methods more explicit. The analysis of the additive and multiplicative Schwarz iterations can be carried out in almost the same spirit as in the traditional block-matrix situation, making convergence proofs of multilevel and domain decomposition methods clearer, or, at least, more classical. In addition, we present a new bound for the convergence rate of the appropriately scaled multiplicative Schwarz method directly in terms of the condition number of the corresponding additive Schwarz operator. These results may be viewed as an appendix to the recent surveys [X], [Ys].

An unconditionally stable finite element time-domain solution of the vector wave equation

Abstract: This paper presents an implicit finite element time-domain (FETD) solution of the time-dependent vector wave equation. The time-dependent formulation employs a time-integration method based on the Newmark-Beta method. A stability analysis is presented demonstrating that this leads to an unconditionally stable solution of the time-dependent vector wave equation. The advantage of this formulation is that the time step is no longer governed by the spatial discretization of the mesh, but rather by the spectral content of the time-dependent signal. A numerical example of a three-dimensional cavity resonator is presented studying the effects of the Newmark-beta parameters on the solution error. Optimal choices of parameters are derived based on this example. >

A priori tests of large eddy simulation of the compressible plane mixing layer

Abstract: Three important aspects for the assessment of the possibilities of Large Eddy Simulation (LES) of compressible flow are investigated. In particular the magnitude of all subgrid-terms, the role of the discretization errors and the correlation of the turbulent stress tensor with several subgrid-models are studied. The basis of the investigation is a Direct Numerical Simulation (DNS) of the two- and three-dimensional compressible mixing layer, using a finite volume method on a sufficiently fine grid. With respect to the first aspect, the exact filtered Navier-Stokes equations are derived and all terms are classified according to their order of magnitude. It is found that the pressure dilatation subgrid-term in the filtered energy equation, which is usually neglected in the modelling-practice, is as large as e.g. the pressure velocity subgrid-term, which in general is modelled. The second aspect yields the result that second- and fourth-order accurate spatial discretization methods give rise to discretization errors which are larger than the corresponding subgrid-terms, if the ratio between the filter width and the grid-spacing is close to one. Even if an exact representation for the subgrid-scale contributions is assumed, LES performed on a (considerably) coarser grid than required for a DNS, is accurate only if this ratio is sufficiently larger than one. Finally the well-known turbulent stress tensor is investigated in more detail. A priori tests of subgrid-models for this tensor yield poor correlations for Smagorinsky's model, which is purely dissipative, while the non-eddy viscosity models considered here correlate considerably better.

Effect of artificial stress diffusivity on the stability of numerical calculations and the flow dynamics of time-dependent viscoelastic flows

Abstract: In this work, we investigate the effect of the introduction of a stress diffusive term into the classical Oldroyd-B constitutive equation on the numerical stability of time-dependent viscoelastic flow calculations. The channel Poiseuille flow at Re ⪢ 1 and O(1) We is chosen as a test problem. Through a linear stability analysis, we demonstrate that the introduction of a small amount of (dimensionless) diffusivity, typically of the order 10−3, does not affect the critical eigenmodes of the viscoelastic Orr-Sommerfeld problem appreciably. However, a diffusive term of that magnitude is shown to have a significant influence on the singular eigenmodes of the classical Oldroyd-B model, associated with the continuum spectra. A finite amplitude perturbation is constructed as a linear superposition of the eigenvectors corresponding to the most unstable eigenvalues of the problem. This is superimposed on the steady Poiseuille flow solution to provide the initial conditions for time-dependent simulations. The numerical algorithm involves a fully spectral spatial discretization and a semi-implicit second order integration in time. For the Oldroyd-B fluid, depending on the magnitude of the initial perturbation, numerical instabilities set in at relatively short times while the components of the conformation tensor increase monotonically in magnitude with time. Introduction of a diffusive term into this model is shown to stabilize the calculations remarkably, and for a three-dimensional simulation with Re = 5000 and We = 1, no instabilities were observed even at very large times. The effect of the magnitude of the diffusivity on the stability and the flow dynamics is addressed through a direct comparison of the results with those obtained for the Oldroyd-B model.

Building Sugeno-type models using fuzzy discretization and orthogonal parameter estimation techniques

Abstract: This paper develops a new approach to building Sugeno-type models. The essential idea is to separate the premise identification from the consequence identification, while these are mutually related in the previous methods. A fuzzy discretization technique is suggested to determine the premise of the model, and an orthogonal estimator is provided to identify the consequence of the model. The orthogonal estimator can provide information about the model structure, or which terms to include in the model, and final parameter estimates in a very simple and efficient manner. The well-known gas furnace data of Box and Jenkins is used to illustrate the proposed modeling approach and to compare its performance with other statistical and fuzzy modeling approaches. It shows that the performance of the new approach compares favorably with these existing techniques.

Comparison of the assumed modes and finite element models for flexible multilink manipulators

Abstract: Compares two discretization models-namely, the assumed modes and finite element models-to efficiently represent the link flexibility of robot manipulators. We present a systematic modeling procedure based on homogeneous transformation matrices for spatial multilink flexible manipulators with both revolute and prismatic joints. The Lagrangian formulation of dynamics and computer algebra are employed to derive closed-form equations of motion. We show that fewer mathematical operations are required for inertia matrix computation in the finite element model compared with the assumed modes formulation; however, because the number of state-space equations is larger, the numerical simulation time may be greater for finite element models. Use of the finite element model to approximate flexibility usually gives rise to an overestimated stiffness matrix. We analytically show that overestimation of structure stiffness may lead to unstable closed-loop response of the original manipulator system, using a model-based control law. We illustrate the complexity owing to the time-dependent frequency equation of the assumed modes model arising in a prismatic jointed flexible link with payload and in manipulators with more than one link with revolute joints. We describe a novel method based on the differential form of the frequency equation to simulate such systems. A model-based decoupling control law is used to compare the dynamic responses of the manipulator system. The results are illustrated by numerical simulation of a flexible spatial RRP (revolute/revolute/prismatic) configuration robot.

A space-time multigrid method for parabolic partial differential equations

Abstract: We consider the solution of parabolic partial differential equations (PDEs). In standard time-stepping techniques multigrid can be used as an iterative solver for the elliptic equations arising at each discrete time step. By contrast, the method presented in this paper treats the whole of the space-time problem simultaneously. Thus the multigrid operations of smoothing and coarse-grid correction are defined on all of the space-time variables of a given grid level. The method is characterized by a coarsening strategy with prolongation and restriction operators which depend at each grid level on the degree of anisotropy of the discretization stencil. Numerical results for the one- and two-dimensional heat equations are presented and are shown to agree closely with predictions from Fourier mode analysis.

Seismic response of three-dimensional alluvial valleys for incident P, S, and Rayleigh waves

Abstract: A simplified indirect boundary-element method (BEM) is presented. It is used to compute the seismic response of three-dimensional alluvial valleys under incident P, S , and Rayleigh waves. The method is based on the integral representations for scattered elastic waves using single-layer boundary sources. This approach is called indirect BEM in the literature as the sources strengths should be obtained as an intermediate step. Scattered waves are constructed at the boundaries from which they radiate. Therefore, this method can be regarded as a numerical realization of Huygens' principle. Boundary conditions lead to a system of integral equations for boundary sources. A simplified discretization scheme is used. It is based on the approximate rectification of the surfaces involved using circles for the numerical and analytical integration of the exact Green's function for the unbounded elastic space. Various examples are given for three-dimensional problems of scattering and diffraction of elastic waves by soft elastic inclusion models of alluvial deposits in an elastic half-space. Results are displayed in both frequency and time domains. These results show the significant influence of locally generated surface waves in seismic response, and they evince three-dimensional effects.

Numerical study of secondary flows of viscoelastic fluid in straight pipes by an implicit finite volume method

Abstract: In this paper, a general class of viscoelastic model is used to investigate numerically the pattern and strength of the secondary flows in rectangular pipes as well as the influence of material parameters on them. To solve the coupled governing equation system, an implicit finite volume method based on the SIMPLEST algorithm, which is applicable for both time-dependent and steady-state flow computations, has been developed and extended for viscoelastic flow computations by applying the decoupled techniques. The main feature of the method is to split the solution process into a series of steps in which the continuity of the flow field is enforced by solving a Poisson's equation for the pressure, and at the end of the steps, both the pressure and velocity fields are made to satisfy one and the same momentum equation. For viscoelastic flow computations, artificial diffusion terms are introduced on both sides of the discretized constitutive equations to improve numerical stability. It is found that there are in total two vortices in each quadrant of the pipe at different aspect ratios (from 1 to 16), and at each ratio the pattern of secondary flows takes the same form for different material parameters, but their strength is very sensitive to the viscoelastic material parameters. Numerical results indicate that the presence of secondary flow strongly depends on the primary flow rate and the elasticity of the fluid, namely, the first and the second normal stress differences as well as their functional departure from the constant multiple viscosity.

A 3-d finite-difference algorithm for dc resistivity modelling using conjugate gradient methods

Abstract: SUMMARY An accurate and efficient 3-D finite-difference forward algorithm for DC resistivity modelling is developed. The governing differential equations of the resistivity problem are discretized using central finite differences that are derived by a second-order Taylor series expansion. Electrical conductivity values may be arbitrarily distributed within the half-space. Conductivities at the grid points are calculated by a volume-weighted arithmetic average from conductivities assigned to grid cells. Variable grid spacing is incorporated. The algorithm does not limit the number and configuration of the sources, although all illustrative examples are computed using two current electrodes at the surface. In general, the linear set of equations resulting from this kind of discretization is non-symmetric and requires generalized numerical equation solvers. However, after symmetrizing the matrix equations, the ordinary conjugate gradient method becomes applicable. It takes advantage of the matrix symmetry and, thus, is superior to the generalized methods. An efficient SSOR-preconditioner (SSOR symmetric successive overrelaxation) provides fast convergence by decreasing the spectral condition number of the matrix without using additional memory. Furthermore, a compact storage scheme reduces memory requirements and accelerates mathematical matrix operations. The performance of five different equation solvers is investigated in terms of cpu time. The preconditioned conjugate gradient method (CGPC) is shown to be the most efficient matrix solver and is able to solve large equation systems in moderate times (approximately 21/2 minutes on a DEC alpha workstation for a grid with 50 000 nodes, and 48 minutes for 200000 nodes). The importance of the tolerance value in the stopping criterion for the iteration process is pointed out. In order to investigate the accuracy, the numerical results are compared with analytical or other solutions for three different model classes, yielding maximum deviations of 3.5 per cent or much less for most of the computed values of the apparent resistivity. In conclusion, the presented algorithm provides a powerful and flexible tool for practical application in resistivity modelling.

Wavelets, spectrum analysis and 1/f processes

Abstract: The purpose of this paper is to evidence why wavelet-based estimators are naturally matched to the spectrum analysis of 1/f processes. It is shown how the revisiting of classical spectral estimators from a time-frequency perspective allows to define different wavelet-based generalizations which are proved to be statistically and computationally efficient. Discretization issues (in time and scale) are discussed in some detail, theoretical claims are supported by numerical experiments and the importance of the proposed approach in turbulence studies is underlined.