Showing papers in "Pamm in 2003"
TL;DR: The main characteristics and application of AD are reviewed, the methodology on a simple example is illustrated and the main techniques used are illustrated.
Abstract: Automatic, or algorithmic, differentiation (AD) is a chain rule-based technique for evaluating derivatives of functions given as computer programs for their elimination. We review the main characteristics and application of AD and illustrate the methodology on a simple example.
94 citations
TL;DR: An overview on how TAF approaches typical challenges of AD such as handling of badly written program code, of large memory/disk requirements, of iterative solvers or of black box routines is given.
Abstract: This paper features FastOpt's automatic differentiation (AD) tool Transformation of Algorithms in Fortran (TAF), a source to source translator for programs written in Fortran 77-95. TAF and its predecessor TAMC have a long record of successful large-scale applications. Here we give an overview on how TAF approaches typical challenges of AD such as handling of badly written program code, of large memory/disk requirements, of iterative solvers or of black box routines. We also point out, where the user is required to prepare his program code prior to invoking TAF.
83 citations
TL;DR: A novel simulation procedure for the estimation of structural reliability is outlined, based on a stepwise algorithm which exhibits convergence rates that are independent of the dimensionality of the problem and outperform direct Monte Carlo as well as existing methods.
Abstract: A novel simulation procedure for the estimation of structural reliability is outlined. It is based on a stepwise algorithm which exhibits convergence rates that are independent of the dimensionality of the problem and outperform direct Monte Carlo as well as existing methods ([2]).
55 citations
TL;DR: These two AD tools provide derivatives for programs written in the MATLAB language, which is widely used for prototype and production software in scientific and engineering applications, while ADMIT/ADMAT implements a pure operator overloading approach of AD, ADiMat also employes source transformation techniques.
Abstract: Derivative information is required in numerous applications, including sensitivity analysis and numerical optimization. For simple functions, symbolic differentiation–done either manually or with a computer algebra system–can provide the derivatives, whereas divided differences (DD) have been used traditionally for functions defined by (potentially very complex) computer programs, even if only approximate values can be obtained this way. An alternative approach for such functions is automatic differentiation (AD), yielding exact derivatives at often lower cost than DD, and without restrictions on the program complexity. In this paper we compare the functionality and describe the use of ADMIT/ADMAT and ADiMat. These two AD tools provide derivatives for programs written in the MATLAB language, which is widely used for prototype and production software in scientific and engineering applications. While ADMIT/ADMAT implements a pure operator overloading approach of AD, ADiMat also employes source transformation techniques.
45 citations
TL;DR: A simple direct asymptotic analysis is proposed which is readily applicable to finite difference discretizations of initial boundary value problems in general and to lattice Boltzmann methods in particular.
Abstract: The correct implementation of Navier-Stokes boundary conditions in the framework of lattice Boltzmann schemes is complicated by the non-availability of analytical methods to assess the consistency of such discretizations. To close this gap, we propose a simple direct asymptotic analysis which is readily applicable to finite difference discretizations of initial boundary value problems in general and to lattice Boltzmann methods in particular. Results of the analysis applied to the classical lattice Boltzmann scheme with bounce back boundary condition are reported.
42 citations
TL;DR: Two strategies for the implementation of Automatic Differentiation based on the operator overloading facility in C++ are presented and the capabilities of the AD‐tool ADOL‐C that appliesoperator overloading to differentiate C‐ and C++‐code are described.
Abstract: In this paper, we present two strategies for the implementation of Automatic Differentiation (AD) based on the operator overloading facility in C++. Subsequently, we describe the capabilities of the AD-tool ADOL-C that applies operator overloading to differentiate C- and C++-code. Finally, we discuss some applications of ADOL-C.
40 citations
TL;DR: In this article, the existence of traveling wave solutions to the Keller-Segel model was studied and various functionals modeling the reactive feedback on the chemoattractant do allow for traveling waves and a wide range of qualitatively different behavior is possible.
Abstract: In this paper we study the existence of traveling wave solutions to the Keller-Segel model, a general model of chemotaxis, where the species do not reproduce. In the case of logarithmic sensitivity we show that various functionals modeling the reactive feedback on the chemo-attractant do allow for traveling waves and a wide range of qualitatively different behavior is possible. We can find monotone fronts as well as pulse solutions in the densities of the population and the chemical. In particular, a new kind of solution exists, where both densities travel as pulses.
35 citations
TL;DR: A general variational based approach for image registration which allows the choice of a user supplied similarity measure and a users supplied regularizer as well as the integration of external knowledge, like, for example, the location of outstanding points, is presented.
Abstract: One of the challenging tasks in today image processing is image registration. Image registration is inevitable whenever images taken for example at different times or from different perspectives need to be compared or to be integrated. Typically, the location of corresponding points in the different views of one object or even of different objects is distorted. For example, motion or different properties of the underlying optical systems (MR, CT) are responsible for the distortion. Thus, a basic problem is to find a meaningful spatial transformation of a given image, such that the transformed image becomes similar to a given second one. Typically, the transformation is computed by minimizing a suitable similarity measure. For many applications it is also desirable to guide the registration by additional information, like the locations of outstanding points. In this note, be present a general variational based approach for image registration which allows the choice of a user supplied similarity measure and a user supplied regularizer as well as the integration of external knowledge, like, for example, the location of outstanding points.
31 citations
TL;DR: The solution of nonlinear stochastic partial differential equations is considered by a Galerkin‐method and by projection in the Stochastic dimension and the occurring high‐dimensional integrals are computed by sparse (Smolyak‐ and Monte Carlo‐integration.
Abstract: We consider the solution of nonlinear stochastic partial differential equations by a Galerkin-method and by projection in the stochastic dimension and compute the occurring high-dimensional integrals by sparse (Smolyak)- and Monte Carlo-integration.
24 citations
21 citations
TL;DR: The intpakX package as mentioned in this paper is a Maple package for interval arithmetic, which is based on intpak, an experimental package that was created by AE Connell and RM Corless in 1993.
Abstract: intpakX, a Maple package for interval arithmetic created in 1999 by I Geulig and W Kramer [1,3], is based on intpak, an experimental Maple package that was created by AE Connell and RM Corless in 1993 [2]
intpakX contains basic data types and operators for interval arithmetic as well as a variety of numerical methods using intervals It offers the possibility to compute verified numerical results with a Computer Algebra System and at the same time display the results graphically
With the new version intpakX v10, intpakX has been updated and redesigned to work with Maple7 This paper shows the range of features of intpakX and gives some examples of their use
TL;DR: A hierarchical multilevel preconditioner is constructed for an efficient solution of a first kind boundary integral equation with the single layer potential operator discretized by a boundary element method.
Abstract: A hierarchical multilevel preconditioner is constructed for an efficient solution of a first kind boundary integral equation with the single layer potential operator discretized by a boundary element method. This technique is based on a hierarchical clustering of all boundary elements as used in fast boundary element methods. This hierarchy is applied to define a sequence of nested boundary element spaces of piecewise constant basis functions as used in the definition of the preconditioning multilevel operator.
TL;DR: In this article, a variational approach is used to find the shortest curves confined to lie in two orthogonal planes and separated by a constant distance, which is applicable to constructing tight shapes of linked structures each component of which is known to be planar.
Abstract: A variational approach is used to find the shortest curves confined to lie in two orthogonal planes and separated by a constant distance. The method is applicable to constructing tight shapes of linked structures each component of which is known to be planar. The shapes of the Borromean rings and two clasped pieces of rope are two examples. A concept of tight periodic structures is introduced and discussed.
TL;DR: In this paper, a stable implicit first-order time-stepping method for simulation of stiff mechanical systems with unilateral constraints and Coulomb friction is presented, which ensures that the unilateral constraints are fulfilled directly on the displacement level.
Abstract: This paper presents a stable implicit first order time-stepping method for the simulation of stiff mechanical systems with unilateral constraints and Coulomb friction. It ensures that the unilateral constraints are fulfilled directly on the displacement level. The resulting linear complementarity problem is formulated in a very compact nonstandard way. A modified form of Lemke's algorithm is presented to solve it.
TL;DR: In this article, the authors considered the dynamic aspects of the combined satellite system and the mechanical stresses in the slender boom and the robotic joints and arm structure to find the optimal capturing strategy.
Abstract: Since several years the interest in offering satellite services for maintenance and repair of malfunctioned satellites is strongly increasing. Currently, a very typical and actual application is dedicated to the rescue of a defect satellite (Rosat) in low earth orbit by means of a servicer satellite based manipulator [1]. The dynamic aspects are carefully considered in order to minimize both, the strong vibrational behavior of the combined satellite system, and the mechanical stresses in the slender boom and the robotic joints and arm structure, and consequently, to find the optimal capturing strategy.
TL;DR: The boundary value problems of the steady oscillations of the linear theory of thermoelasticity with microtemperatures are investigated in this article, and the uniqueness and existence theorems of solution of the boundary value problem by means of boundary integral method are proved.
Abstract: The boundary value problems of the steady oscillations of the linear theory of thermoelasticity with microtemperatures are investigated. The uniqueness and existence theorems of solution of the boundary value problems by means of the boundary integral method (potential method) are proved. The Sommerfeld-Kupradze type radiation conditions are established. The existence of eigenfrequencies of the interior homogeneous boundary value problems of steady oscillations is studied.
TL;DR: In this paper, a stress concentration problem in a circular cylindrical vessel with ellipsoidal heads was analyzed and the optimal head was determined on the grounds of three optimisation criteria.
Abstract: The subject of analysis is a stress concentration problem in a circular cylindrical vessel with ellipsoidal heads. The region of the junction of ellipsoidal and cylindrical shells is subject to stress concentration. Intensity of the concentration depends, first of all, on relative convexity of the ellipsoidal head, and the ratio of thicknesses of the shells. Based on the boundary disturbance theory the stress state in this cylindrical vessel was analytically described. Optimal ellipsoidal heads were determined on the grounds of three optimisation criteria.
TL;DR: In this article, an asymptotic analysis of the stress fields near an unsymmetric bimaterial notch with arbitrary opening angle is performed using the complex potential method, and the order of the singularity at a notch tip is determined in closed-form analytical manner, so that the dependency of the occurring singularity exponents on geometry and material properties can be studied systematically.
Abstract: An important issue in the mechanics of adhesive bonds is the knowledge of local mechanical fields. In the present study, an asymptotic analysis of the stress fields near an unsymmetric bimaterial notch with arbitrary opening angle is performed. Using the complex potential method, the order of the singularity of the stress fields at a notch tip can be determined in closed-form analytical manner, so that the dependency of the occurring singularity exponents on geometry and material properties can be studied systematically.
TL;DR: A stochastic partial differential equation is discretized by stoChastic spectral finite elements and the resulting large linear system is solved by parallel iterative solvers, exploiting various forms of parallelism.
Abstract: A stochastic partial differential equation is discretized by stochastic spectral finite elements. The resulting large linear system is solved by parallel iterative solvers, exploiting various forms of parallelism.
TL;DR: Three different types of elastic registration approaches which have different properties and thus differ w.r.t. their suitability for certain registration applications are developed and all approaches are both computationally efficient and can be applied for multimodal images.
Abstract: For the geometric alignment of image data we propose spline-based schemes utilizing anatomical point landmarks. We have developed three different types of elastic registration approaches which have different properties and thus differ w.r.t. their suitability for certain registration applications. All approaches are both computationally efficient and can be applied for multimodal images. Our approaches have been successfully applied for the registration of 2D and 3D tomographic images of the human brain.
TL;DR: In this article, the performance and the efficiency of smart structures controlled by piezoelectric wafers decisively depend on the locations of the actuators and sensors, and the results of a typical example are presented.
Abstract: The performance and the efficiency of smart structures controlled by piezoelectric wafers decisively depend on the locations of the actuators and sensors. For the evaluation and testing of time consuming numerical optimisation techniques, which are currently a part of our research, benchmark examples are required. As such a benchmark example a simply supported plate structure is inverstigated, where the active controlled behaviour can still be described analytically. In the paper the optimisation algorithm is briefly described and the results of a typical example are presented.
TL;DR: In this article, an experimental investigation of the flow around a finite circular cylinder mounted on a flat plate is reported, where complex 3D fluid motions in the separated flow region are induced by the superposition of three main vortical flows.
Abstract: An experimental investigation of the flow around a finite circular cylinder mounted on a flat plate, see fig. 1, is reported. The aspect ratio L/D (with length L and diameter D) of the cylinder model is 2.0. The focus of this study is toward examining the complex separated flow structures and wake properties. Velocity and turbulence measurements have been carried out with a three component Laser Doppler anemometer (LDA) at the Reynolds number ReD = 2.0 · 105. The experimental results show complex 3D fluid motions in the separated flow region. They are induced by the superposition of three main vortical flows.
TL;DR: In this article, the motion of liquid drops on horizontal surfaces resulting from a contact angle gradient was analyzed using a video of the initial diameters of 0.31 - 0.5 mm (15 - 65 nl).
Abstract: Experimental results on the motion of liquid drops on horizontal surfaces resulting from a contact angle gradient are presented. Silicon surfaces were modified using dodecyltrichlorosilane to generate the gradient. Water drops with initial diameters of 0.31 - 0.5 mm (15 - 65 nl) were placed on the surface, their movement videotaped, and subsequently analyzed. To characterize the gradient surface the static contact angle was measured along the surface.
TL;DR: In this article, a variational principle for a poroelastic thin plate model is developed. But the principle is not applicable to the case of thin plate models, since the pore pressure and the solid displacements are unknown.
Abstract: The numerical treatment of noise insulation of solid walls has been an object of scientific research for many years. The main noise source is the bending vibration of the walls usually modeled as plates. Generally, walls consist of porous material, for instance concrete or bricks. Therefore, a poroelastic plate theory is necessary. A theory of dynamic poroelasticity was developed by Biot using the solid displacements and the pore pressure as unknowns. After formulating the poroelastic theory for thick plates, Mindlin's theory, a variational principle for this poroelastic thick plate model is developed. This is the basis of a Finite Element formulation.
TL;DR: In this paper, a quasi-static state feedback is proposed that achieves stable tracking for planar motions of rigid bodies and is invariant with respect to the choice of the inertial frame.
Abstract: A quasi-static state feedback is proposed that achieves stable tracking for planar motions of rigid bodies and is invariant with respect to the choice of the inertial frame.
TL;DR: A global density gradient based regularization is proposed and the results of numerical studies of several maximum‐stiffnessdesign problems based on the presented approach are discussed.
Abstract: The formulation of topology optimization problems for continuum structures usually requires the introduction of the material density as design variable. In this context the design domain is discretized by finite elements and a constant density distribution within the elements is assumed. To prevent the occurrence of optimal designs characterized by extended regions with intermediate density values, special penalty methods are used. The introduction of those methods results in discontinuities in the global density distribution and often leads to designs containing unfavourable microstructures such as the well-known checkerboard patterns. To obtain designs free of microstructures, we propose a global density gradient based regularization and discuss the results of numerical studies of several maximum-stiffnessdesign problems based on the presented approach.
TL;DR: In this article, the experimentelle Verifizierung der Beschreibung and Ansatze zur Identifikation der Zeitkonstanten anhand von Geschwindigkeitssprungen oder -rampen is discussed.
Abstract: Die schnelle Geschwindigkeitsanderung wahrend einer Stick-Slip-Schwingung beeinflust oft die Reibkrafte im Kontakt so stark, dass eine einfache Beschreibung mit einer stationaren Reibkraftcharakteristik nicht mehr sinnvoll ist. Die instationaren Reibkrafte konnen mit einer Differentialgleichung anhand der stationaren Reibkraftcharakteristik und zweier Zeitkonstanten beschrieben werden. Schwerpunkt dieser Untersuchung ist die experimentelle Verifizierung der Beschreibung sowie Ansatze zur Identifikation der Zeitkonstanten anhand von Geschwindigkeitssprungen oder -rampen.
TL;DR: In this paper, numerical simulations of separating and reattaching turbulent boundary layer flow over a backward-facing step are analyzed with respect to the time-dependent flow behaviour, and the effects of open-loop passive and active flow control methods are investigated.
Abstract: Results from numerical simulations of separating and reattaching turbulent boundary layer flow over a backwardfacing step are analysed with respect to the time-dependent flow behaviour. Beside the well-known roll-up of the separated shear-layer (“Kelvin-Helmholtz instability”) and the ejection of subsequently formed large-scale structures (denoted as “shedding”), another unsteady phenomenon which is commonly called “flapping” can be observed. Additionally, the effects of open-loop passive and active flow control methods are investigated.
TL;DR: In this article, the authors present a probabilistic description of abrasion and crush dressing of grinding wheels and derive relations describing distributions of active grain age and its change during of grinding.
Abstract: The aim of the paper is to present modelling of abrasive wear and grinding wheel durability by means of probabilistic description of abrasion and crush dressing of grinding wheels. Relations describing distributions of active grain age and its change during of grinding are determined. A practical application of the models proposed in the paper, using empirical data, makes it possible to determine other quantities which are very difficult to estimate in an experimental way.
TL;DR: In this article, the onset of the dynamic instability in a.oating caliper disc brake system is studied, where the disc is modelled as a.exible rotating plate, the pad is taken as a small mass and a distributed sti.ness, and the motion of the caliper is also taken into consideration.
Abstract: In this paper, the onset of the dynamic instability in a .oating caliper disc brake system is studied. The disc is modelled as a .exible rotating plate, the pad is taken as a small mass and a distributed sti.ness, and the motion of the caliper is also taken into consideration. The linearized equations of motion about the equilibrium position are derived assuming a constant braking force. The equations are subsequently discretized with the help of appropriate shape functions. The eigenvalue problem is then solved to detect the onset of instability. The effects of damping on the stability of motion are investigated as well.