In the present paper, incompressible hyperelastic materials are analyzed using GSMAC-FEM that is a kind of decoupled FEM. Higher-order elements are applied to GSMAC-FEM, and stiffening property of rubber undergoing large deformation is introduced in order to use this computational method in practical problems. The accuracy of simple deformations is verified by using various elements in two-dimensional or three-dimensional problems. Then, stress-strain curves like S-letter are made by calculating the element-average values of elastic constants of higher-order Mooney-Rivlin material. Finally, the differences between various elements are compared in terms of the accuracy of solutions under the severe condition in compression of block problem that has singularity.
An analysis of spatio-temporal pattern created by self-organization is one of important problems in complex systems. Among many spatio-temporal patterns created by self-organization, focusing on chemotactic bacterial colony patterns formed by Escherichia coli (E. coli) and Salmonella typhimurium (S. typhimurium), the colony formation processes are analyzed. The bacteria E. coli and S. typhimurium move towards higher concentration of chemoattractant or lower concentration of chemorepellent. This property of bacteria is called chemotaxis. Taking the fluctuation such as impurities in chemical substance into consideration, the influence of the fluctuation on the bacterial colony formations is studied by numerical simulations.
Concrete reinforced with short steel fibers is associated with its ability to control cracking and enhances the tensile properties of concrete. For this reason, Fiber reinforced concrete has been widely used in infrastructures, where tensile cracks may occur. This research is a numerical study to predict the mechanical properties of steel fiber reinforced concrete SFRC using the finite element method as an alternative method instead of the costy material experimental tests. The numerical approach helps study the effect of the random distribution of fibers on the mechanical properties of SFRC. The flexural and energy absorption are investigated using different fiber contents. The growth of cracks and damaged elements can be predicted. The paper also aims to study the influence of fibers on creep shrinkage of fiber reinforced concrete and the results are agreeable with the experiment results.
The present paper deals with a truss topology optimization problem whose objective functions are both structural volume and compliance. This problem can be formulated as a multiobjective optimization problem (MOP), which is to find a set of cross-sectional area of members, such that structural volume and compliance should be minimized. In this paper, the Pareto optimal front of the present MOP is theoretically obtained using the Kuhn-Tucker conditions for MOPs. Moreover, based upon the characteristics of the theoretical Pareto front, the present MOP of a truss structure is transformed into an unconstrained minimization problem of the product of structural volume and compliance. A genetic algorithm is applied to this single objective optimization problem in this paper. The application of the proposed method is illustrated in numerical examples with discussion on both efficiency and accuracy.
This paper proposes JSGrid, which is a software environment to build up computer clusters running on various operating systems. JSGrid makes good use of idle computational resources without installing any software. JSGrid therefore enables various kinds of computers to be a part of a cluster, for instance, educational terminals at schools and universities, computers for development, office computers at companies and laboratories, and information retrieval terminals at libraries. Experimental results have shown that JSGrid could manage various operating systems such as Windows, Linux, and Mac OS, and keep high throughput almost equal to theoretical prediction even under a circumstance in which dynamic join and termination of computers occur.
Accurate and consistent methods are required to reconstruct the surface of an object from a set of points, which are located on the surface of the object. The detection and generation of feature lines appearing on the surface is considered to be an attractive preprocessing step before the surface reconstruction process. In this paper, the authors propose a method to detect feature lines from point clouds. The detection and generation process depends on a number of effective parameters. The authors have identified four effective parameters for feature line generation: 1) the number k nearest neighbors, 2) curvature value, 3) cycle length and 4) minimum length of short branches. The efficiency of the detection by finding the parameter values that closely match the existing feature lines are examined through test problems of point clouds. Finally, the effectiveness of the method is evaluated by application to a remote sensed point data set representing topographic features of the Jordan valley.
We consider a preconditioner of the MRTR method for solving a linear system of equations with real symmetric positive definite matrix. The proposed utilization of preconditioner has an advantage of preserving symmetric positive definiteness. The residual vector of MRTR method is the product of the starting residual vector γ0 and Hκ(A) only. Moreover the MRTR method is also based on the three-term recurrence formular of the CG method. Numerical experiments show effective convergence rate of the proposed preconditioning of the MRTR method as compared with the existing preconditioner.
A CIVA-lattice Boltzmann method (CIVA-LBM) for incompressible viscous flow simulation using unstructured triangular mesh is proposed. The cubic interpolation scheme based on CIVA is employed to solve the streaming step of the discrete Boltzmann equation. In this paper, the basic idea of the CIVA-LBM is presented first, and the solution procedure of the collision step of the dicsrete boltzmann equation and its accuracy and stability are described. The numerical performance of the CIVA-LBM is tested on the analysis of the square driven cavity flow and the flow past a circular cylinder, and the numerical accuracy and the stability of the method is investigated.
The framework system which enables scientists and engineers to do interdisciplinary studies of computational science and engineering is developed using open source software. The system is a web-based system which is composed of problem solving environment user interface and collaborative research environment. The framework system is applied to particle image velocimetry research of fluid, genome analysis using basic alignment search tool, and mobile communication system design. The framework system is an easy-to-use and powerful system for interdisciplinary studies.
We choose s-version FEM as a useful tool for analysing of composite materials. The drawback of s-version FEM is to make it difficult to solve a linear system of equations which stems from discretization by s-version FEM. Therefore we consider how to efficiently solve the linear systems. In this paper we describe an efficient implementation of matrix-vector computation, and propose a masking preconditioner of the CG method. Through numerical experiments we verify effectiveness of the proposed masking preconditioner suited to a rich variety of composite materials analysis.
A distributed computing middleware named EnGrid is developed to obtain the large computing power from the idling personal computers (PCs) in office environments. Transmission control protocol (TCP) is used for the communication among EnGrid, client program, main job, and sub job. The EnGrid does not request a central server and can be easily installed at office environments. The problem solving environment, whose subsystems are parameter sweep system and parameter optimization system, is developed for determining the parameter for finite element method (FEM). The developed technique was found to be useful when applying it to the analysis of an excavation problem involving retaining wall, using the 2-dimensional civil engineering FEM.
In this study, first, we obtain the elastic modului of the overall composite material based on the Eshelby/Mori-Tanaka model in the case where anisotropic ellipsoidal inhomogeneities with three-dimensional orientations are regularly distributed in an isotropic homogeneous matrix. Then, the elastic moduli of the overall a composite material are determined when anisotropic ellipsoidal inhomogeneities are distributed randomly based on Reuss model and an equivalent model using the concept of local regions in the previous paper. Furthermore, using the analytical equations obtained, concrete numerical examples under the assumption of a ceramic matrix composite material (SiO2/Al2O3) are studied. As a result, we have succeeded in the theoretical determination of the elastic moduli of the overall composite material containing anisotropic ellipsoidal inhomogeneities of arbitrary volume fraction and shape.
Due to the accuracy and the efficiency of hexahedral elements, the all-hexahedral element auto meshing has a growing demand for the finite element analysis. The authors have been studying whisker-weaving algorithm to generate a hexahedral mesh (HM) automatically(1). In this method the prerequisite for generating a HM is a quadrilateral surface mesh (SM). From the given SM, combinatorial dual cycles (whisker sheet loops for the whisker-weaving algorithm) are generated to produce a HM. Generating a good quality HM does not depend only on the quality of quadrilaterals of the SM but also on the quality of the dual cycles generated from it. If the dual cycles have self-intersections, it could cause the formation of degenerated hexahedrons called knife elements, which are not usable in the finite element analysis. The presence of self-intersections in dual cycles depends on the SM. In this paper a detailed method is proposed to generate a HM free of knife elements, which includes modification of the SM to remove self-intersections from its dual loops. The SM modification procedure of the proposed method has three basic steps. These steps are (a) face collapsing, (b) new face generation and (c) template application.
The finite cover method (FCM), which is known as one of the generalized versions of the finite element methods (FEM), and enjoys so-called generalized elements that partially have physical domains. We examine the finite cover approximations for imposing interface compatibility conditions in the generalized elements by introducing a sort of patch tests for a finite cover mesh with an interface. First, we present the formulation of the FCM together with three kinds of interface connection methods for generalized elements; the penalty method (PM), the Lagrange multiplier method and the discontinuous Galerkin method (DGM). Secondly, the patch tests specialized to the FCM are carried out to assess the accuracy of finite cover approximations with respect to both the magnitude of penalty parameters in PM and DGM and the order of interpolation functions for the Lagrange multipliers.
In recent years, many structural and fluid analysis programs are parallelized. The authors have been developing fluid analysis programs using MPS method(Moving Particle Semi-implicit Method). In this study, the authors investigate the proportions of the processes in the MPS code. Based on the above investigation, a parallel solver with variables of the distributed coefficient matrix is developed. In order to evaluate the performance of a parallel program, the parallel efficiency of solver and parallel efficiency of searching particles function are measured using test calculations. One million particles are used for the analyses of water collapse and tsunami to demonstrate the performance of a large scale problem.
A formulation of novel in-plane generalized finite element which can be utilized in geometrical non-linear analysis is presented and some analyses using the proposed element are conducted. The proposed element can reproduce quadratic deformation mode with only corner nodes and it has no linear dependency which is one of the well known problems of generalized finite element. The formulation is based on the rate form of the virtual work principle and obtained by a simple extension of standard FEM. The convergence of analysis solution and robustness for element distortion are investigated and the results are compared with those of standard displacement based first and second order elements. The proposed element provides a good solution convergence which is as well or better than those of the conventional second order elements. Additionally, it is shown that high-precision solution is given when the mesh is strongly distorted.
Medullary bone formed in estrogen-injected male quail has microscopic heterogeneity due to complex morphology of trabeculae and nanoscopic characteristics. The latter arise from alignment of biological apatite (BAp) crystallites and collagen fibrils. Hence, the mechanical functions of bones are not always directly related to BMD (Bone Mineral Density) but to the morphology of medullary bone as well as BAp orientation. In this paper, multi-scale modelling of the above-mentioned medullary bone is presented using the homogenization technique. Its verification by very large-scale FEM analysis using the Earth Simulator is also described. Mesh superposition method is employed for multi-scale in-vivo stress analysis under arbitrary disturbances.
In this study, we present a new numerical scheme for incompressible MHD flow. Finite edge-element method for electromagnetic field is coupled with improved GSMAC-FEM for velocity field weakly on this scheme. By using finite edge-element method, we can analyze the external region together with MHD flow region. Including the non-linear (advection) term of electromagnetic field, this scheme can deal with the interaction between electromagnetic and velocity fields. There are few bench mark problems, therefore, MHD lid-driven cavity flow is adopted for a bench mark problem, and calculated by this scheme. The result is evaluated, and the validity is confirmed from the view point of magnetohydrodynamics.
In the present paper, the explicit computational method for fluid-solid coupled system is proposed based on ALE (Arbitrary Lagrangian-Eulerian) GSMAC(Generalized-Simplified Marker and Cell)-FEM suitable for low capacity and high speed computation in order to analyze the interaction of fluid and flexible rubber-like solid. The stability of time-marching is improved by satisfying the fluid-solid strongly coupled equation in next time using the explicit iterative calculation. Only the Poisson equation for the pressure correction is computed implicitly in this coupled method. In order to verify the effectiveness for the problem that the fluid-solid interaction is large, two-dimensional vortex-induced vibration problem is analyzed about the interaction of vortices and elastic plate attached to rigid prism.
Hybrid-type penalty method (HPM) assumes the linear displacement field with rigid displacement, rigid rotation and constant strain in each sub-domain and introduce subsidiary condition about the continuity of displacement into the framework of the variational expression with Lagrange multipliers. This compatibility of the displacement on the intersection boundary is approximately introduced using the penalty as a spring constant which is applied to the Lagrange multiplier. Present method can deal with the fracture on the intersection boundary and yielding in the each element at the same time. This paper shows the upper and lower bound solution for limit analysis using present method.
Over the past few years, studies have been performed on filter generation methods for image processing with genetic algorithm (GA) or genetic programming (GP). The existing methods generate various filters by combining existing primitive filters. Although filters generated by GP show better output quality than filters generated by GA, chromosome size becomes huge in GP. Uncontrolled chromosome growth, called bloat, makes filter application time worse, and makes it impossible for users to analyze and modify generated filters. This paper proposes a GP based image filter generation method that tunes numeric parameters of the primitive filters and restrains surplus chromosome growth. Experiments with traffic sign extraction problem showed the filters generated by the proposed method had almost the same output quality as filters generated by preceding GP with almost the same size of filters generated by GA.
This paper discusses a variant of the Crout version of Incomplete LU (ILUC) decomposition. Performance of the ILUC decomposition has been reported in the field of earth resource engineering and electromagnetic analysis. We embody a particular strategy for stable convergence of ILUC decomposition by means of compensation of a diagonal entry for fill-in. Numerical experiments indicate that the iterative methods with ILUC decomposition based on compensation of a diagonal entry perform well in view of efficiency, safety of convergence and necessary number of fill-ins as compared with the conventional ILUC decomposition.
When we voxelize a polygon model, we have to check all the voxels in the computational domain whether they are inside or outside of the model. A large number of voxels makes the voxelization cost high. In recent years, the number of vertices in a polygon model is increasing. So efficient voxelization from a large polygon model is a challenging topic. In this paper, a realtime solid voxelization method is proposed. Our method voxelizes a polygon model with tens of thousands of vertices to voxel data less than one second. The polygon model is converted into a set of layered depth images which are independent of model complexity. At the same time, the orientation of polygons are also extracted. In general, a ray traversal method is used in voxelization phase, which counts the number of times hitting a polygon to determine voxel properties. We propose a short ray traversal method, which divides a ray to a set of short rays which go through the volume in parallel. Our method needs not to count the number of hitting times and can check the voxel without any global information. Therefore, the computational cost of short ray traversal has no relation to model complexity. We also propose two optimization methods which compute the unnecessary region of the bounding box for skipping some calculations. These techniques accelerate voxelization.
Cartesian grid method has higher availability for product design of actual problems at the early stages of the development process. Simple block approximation of a shape is of great stability at the grid generation. In this case, the arbitrary patterns of wall elements appear in the computational domain where adequate pressure and velocity boundary condition should be employed. In this paper, an efficient and robust implementation of boundary condition, which is build into the discretized scheme of the equations, is proposed. This method effectively eliminates the difficulties of the reference point that has multiple values of pressure and velocity. As a validation, incompressible flows in opposed ducts, which are separated by a single wall element, was calculated using this method. Comparing with the simple but inaccurate implementation of boundary conditions, the proposed method shows more accurate results. In addition, several geometrically severe test cases are performed for single cell obstacle and narrow flow channel. Finally, it was found that present method has desirable properties to solve flows around complex geometries as a first order analysis.
In case of the heteroepitaxial growth of a GaAs thin film on a Si substrate, one of the technical problems is increase of dislocation density in the GaAs thin film during the cooling process from growth temperature to room temperature. Such increase of dislocation density is caused by the difference of thermal expansion coefficients between GaAs and Si. In this study, the dislocation density in the GaAs layer during the cooling process was numerically analyzed by the three-dimensional finite element method. A dislocation kinetics model called the Haasen-Alexander-Sumino model was used as a constitutive equation for creep. Cooling velocity and thickness of the GaAs layer were changed as parameters, and their effect on the dislocation density was evaluated. The effect of the rapid thermal annealing on the dislocation density was also discussed.
When applying SA to continuous optimization problems, the appropriate adjustment of the neighborhood ranges becomes necessary to obtain the good performance. In this paper, we propose a Neighborhood Parallel Simulated Annealing (NPSA) for continuous optimization problems, which provides global search using the periodic exchange of different neighborhood ranges with parallel computers. The proposed approach, NPSA, automatically determines the appropriate neighborhood range, and shows a good perforamance in solving typical test problems.
In order to improve the convergence property of the preconditioned conjugate gradient (PCG) method for solving eigenvalue problems of the Hamiltonian matrix, we propose a new preconditioning method. The preconditioner utilizes not only an approximate eigenvalue which is obtained during the CG iterations but also its residual error. We demonstrate that the PCG method with the new preconditioner can solve the eigenvalue problem for the Hamiltonian matrix several times faster than the PCG method with the conventional preconditioner.
Ingot annealing process is indispensable for compound semiconductor single crystals such as GaAs and InP to improve their electric characteristics. One of the technical problems of ingot annealing is increase of dislocation density that affects the performance of electronic devices. A computer code was developed for the dislocation density evaluation of a single crystal during ingot annealing process. A dislocation kinetics model called the Haasen-Alexander-Sumino model was used as a constitutive equation. In this model, creep strain rate is related to the dislocation density, and this model was extended to the multiaxial stress state based on the theory of crystal plasticity. Three-dimensional finite element model was used to take account of crystal anisotropy in elastic constants and specific slip directions. The authors improve the creep constitutive equation to take account of dislocation annihilation. Dislocation density analyses during ingot annealing process were performed for an InP single crystal with 101.6mm in diameter. The distribution and time variation of dislocation density were obtained from this computer code.
The finite cover method (FCM) for quasi-brittle heterogeneous solids and the associated modeling method are developed. First, we introduce the FCM in conjunction with the mortar approximations for material interfaces, and attract attention to the modeling of surfaces of strong discontinuities and material interfaces by means of spatially fixed regular mesh. Secondly, the associated modeling method for complex heterogeneous solids is presented, and the issues in computational implementation are discussed especially for dealing with multiple cohesive crack growth in meso-scale heterogeneous media. Finally, after carrying out simple performance studies, we demonstrate the promise and potential of the proposed method in several numerical examples.
This paper describes the GPU-assisted simulation of elastic wave propagation in solid based on the finite element method. The GPU, Graphics Processing Unit, was first developed only for processing 3D computer graphics several years ago. It has obtained both flexible programmability and fast numerical processing ability. This has led to the so-called GPGPU, General Purpose GPU i.e. applications of GPUs into non-graphics fields. In this paper, GPU is utilized for the simulation of elastic wave propagation in solid based on the dynamic explicit FEM. In the proposed method time integration process is performed on a GPU. The fundamental formulation of the method is described in detail, and its basic performance is tested through sample analyses. Finally, a new efficient algorithm for matrix-vector product is proposed.
A newly designed framework was developed to support the development of efficient systems and to provide execution environment with high functionality for various physical simulation. This framework named SPHERE provides many functions like coupling mechanism between different solvers, parameter description by XML, assistance of parallelization, run-time environment, mixed-language development, and so on. Although this system is described by C++ to receive benefit from the Object-Oriented Programming (OOP) philosophy, a programming is based on a hybrid style that combines object-oriented and conventional procedural languages to achieve high performance and to retain higher portability for existence simulation codes. It was confirmed that the amount of source code was significantly reduced according to the intrinsic mechanism of inheritance of OOP. In addition, comparison of measured timing between an original CFD code written by FORTRAN90 and a ported CFD code on the framework revealed us that SPHERE system showed excellent performance of over 90 percent of original code on several different platforms.
The shallow water equations generate short-wave spurious oscillations if they are solved with the standard Galerkin finite element approach. In this article, a mixed finite element method with the quasi bubble-function element is applied to the equations to solve tidal currents. Numerical solutions of the quasi bubble-function scheme are investigated in comparison with analytical solutions and numerical solutions of the selective lumping method. The results indicate that the quasi bubble-function scheme is stable in the tested subcritical flows and yields accurate solutions with very low numerical dumping. The results of tidal current computations in Tokyo Bay show that the quasi-bubble mixed finite element method reproduces very detailed residual currents near coasts, which is important to transportation analyses, thanks to its low numerical damping property.
It has been problematic to find the directions of bifurcating paths with reference to the critical eigenvectors when the multiplicity of criticality is large. Stiffness modification method, which slightly modifies tangent stiffness matrices to separate the multiple eigenvalues, is a pertinent means for this problem. However, appropriate modification of stiffness matrices has not been clarified and the critical eigenvectors required for searching bifurcation paths cannot be obtained by the stiffness modification method. In this context, through a set of bifurcation analyses of a periodic honeycomb structure, we examine an appropriate modification of the stiffness matrix and how to obtain a set of critical eigenvectors with certain symmetries corresponding to directions of the bifurcating paths.
With increase of the interdisciplinary collaboration between remotely located researchers, a communication environment with high reality becomes essential as a collaboration support tool. In a VizGrid project, we propose a concept called Volume Communication as a new collaboration support tool which provides remote conference system with high reality. In the Volume Communication concept we setup a virtual collaboration room and allocate three-dimensional (3D) video human images of researchers at remote locations, 3D numerical simulation results and measured data. By visualizing the virtual collaboration room from each researchers view point, we can realize an environment as if these researchers communicate each other in the same meeting room. To realize the remote conference system with high reality, we develop voxel communication method which handle 3D video human images, 3D simulation result graphic images and 3D measured data graphic images as voxel data and transmit, allocate, visualize and operate these data. In this paper we report the voxel communication method and evaluated results of its functionality and performance by experimental system.
First, for growing cracks in elastic-plastic materials, an incremental variational principle is developed to satisfy the boundary conditions near newly created crack surfaces. Then using this variational principle, the moving finite element method is formulated and developed, based on the Delaunay automatic triangulation. Furthermore, theoretical backgrounds on numerical prediction for fracture path of curving crack using T* integral are explained. Using the automatic moving finite element method, fracture-path prediction simulation is carried out for elastic-plastic fracture under mixed-mode loading.