This paper proposes a planning and scheduling method for rearrangement of floor layouts (RAFL)-a search problem involving multi-path planning and resource allocation. RAFL consists of pre- and postchanged layouts of objects, workers and a work time limit. The goal in solving RAFL is to find a pair of a rearrangement plan - a set of paths for each object - and its schedule - an assignment of actions for each worker in each turn in order to make objects move along the planned paths. RAFL is difficult to find even a feasible solution due to frequent collision and scrambles for workers among objects. The proposed method represents the problem by a set of various, complex constraints necessary to planning and scheduling, and finds a solution efficiently by using user’s advice as both constraints and heuristics for pruning wasteful search space and weighting significant values and variables. Experimental results showed user’s advice considerably reduced the time to solve and enabled to get a feasible solution within practical time.
This paper presents two techniques, called the morphing method and the fitting method, for deforming finite-element models to conform to prescribed boundary shapes of domains. The morphing method involves selecting nodes on a finite-element model and moving them to desired locations under suitable constraints in reference to specified image data. The fitting method involves deforming a finite-element model to fit into a prescribed domain that can be given with pixel or voxel image data. These methods have been derived as solutions to shape optimization problems for morphing and fitting using the traction method. The morphing method can be applied to construct a finite-element model of the spine to match that of patients with idiopathic scoliosis by referring to X-ray photographs. The fitting method can be applied to construct a skeletal part that matches CT image data.
Large deformation analyses of rods, which are flexible structures, encounter numerical difficulties such as instability or bifurcation. In this work, an arbitrary Lagrangian-Eulerian (ALE) finite element method is applied to elastic rod for the purpose of overcoming such difficulties. The proposed procedure is based on the geometrically exact rod theory. By this procedure, the initial shape can be changed during the large deformation analysis and the solution for the initial shape without imperfection can be calculated from the solution for the imperfect initial shape in the buckling problem.
In this paper, large scale analyses of shell structures by using the CGCG iterative solver are conducted. For more efficient analyses of large scale problems, the demands have been increasing for the efficient iterative solvers and the domain decomposition methods with preconditioners have been developed. In the analyses of thin shells, however, the situation is not so transparent due to the well-known ill conditioned stiffness matrix. Bearing this in mind, the CGCG solver, the conjugate gradient method with coarse grid solver which was already proven to be fast, accurate and robust for 3D solid structures, is adapted for thin shell structures and the performance is compared with those using the other preconditioning techniques based on the domain decomposition methods. As numerical results, the CGCG solver shows better performance from the viewpoints of the convergence rates, the memory usage and the total computing time than the other iterative solvers based on the domain decomposition methods.
Recently, the development of machine tools and sub-micron position-control techniques has brought the minimum thickness of ultra-precision grinding or cutting to less than 1 nm. The conventional finite-element method (FEM) becomes impossible to use for numerical analysis since the focused region and mesh is very small. As another disadvantage of FEM, micro property of the material such as crystalloid was not taken into consideration. As an alternative method, molecular dynamics (MD) method is significantly implemented in the field of micro-cutting, indentation and crack propagation. However, enormous time to calculate the interaction between molecules notably hampered the wide use for large-scale MD simulation. In this paper, firstly, parallel MD based on spatial decomposition and particle decomposition is constructed to reduce the calculation time whose efficiency will be evaluated in supercomputer by benchmark test. Secondly, parallel MD is used to carry out the large-scale nanometric grinding simulation with respect to different condition. By comparison between the simulation results, it is made clear that, variation of the cutting force is dependant on dislocation of the workpiece lattice; the cutting force is approximately equivalent as the wear of tool takes place; and the property of machined surface and temperature distribution is influenced by the cutting speed.
In recent years, the Lattice Boltzmann Method (LBM) has been developed as an alternative numerical approach in Computational Fluid Dynamics (CFD). In particular, this method is promising for simulations of multiphase and multi component fluid flow involving complex interfacial dynamics. Unlike the conventional CFD methods based on NS equations, the LBM is based on the mesoscopic particle’s kinetic equation. This method has some advantages such as the simplicity of the algorithm (high efficiency on parallel processing), flexible reproduction of interfaces between phases. The conventional LBM, however, requires regular structured grids. But, for complex flow field, unstructured grid should be used. In this study, we describe a computational scheme based on two-dimensional unstructured grids using Cubic Interpolation with Volume/Area coordinates (CIVA) method and Adaptive Mesh Refinement method. As examples of tests of this scheme, single and two-phase flow simulations are presented.
This paper presents a parallel finite element analysis of large scale shell structures where the CGCG iterative method is utilized as a linear equation solver. The CGCG method is one of the conjugated gradient methods with the coarse grid solver and has been developed for analysis of shell structures in our previous paper. As the CGCG method is based on the domain decomposition method, it can be adapted to the parallelization straightforwardly. As a numerical example, an analysis of pinched cylinder is conducted. Comparing with the results by the BDD method, the parallel CGCG solver shows competitive performance with regard to the computational time and the parallel efficiency and superior efficiency with regard to memory usage. Finally, a large scale shell problem with 10 million degrees of freedom is successfully solved by the present parallel CGCG solver.
A lot of methodologies for a grid dependency study have been proposed like the analytical method has been developed by Roach. However, a method for practical problems has not been established. Generally, comparison from researchers' viewpoint among the computational result and other solutions (exact solutions, experimental results, etc) is the most popular way to check the grid dependency. In this case, it is important to efficiently collect many results about the same problem, and to compare such various results systematically. In this paper, we propose a concept of the comparison space, and a strategy of grid dependency study using comparison on the Grid.
We propose an efficient searching method of an optimum particle distribution for the minimum porosity in random packing simulation by genetic algorithm (GA). The particle distribution is defined by five central values which are composed of particles diameters and their frequencies, and these parameters have limited upper and lower values. The porosity of the powder heap in the packing mold is calculated and evaluated by random packing simulation algorithm which simulates independent porosity of the packing order. GA evaluates the simulated porosity and searches an optimum particle distribution for minimum porosity as an inverse problem. Our GA simulation has reduced the simulation times to 61% of that of the conventional round-robin method for searching the optimum particle distribution.
Non-blocking communication in the MPI (Message Passing Interface) standard provides us with the functionality to overlap communications with calculations. However, the standards only specify the semantics of the APIs (Application Program Interfaces), and actual implementations only assure the context of communications. In fact, the MPI libraries presently installed on a Compaq AlphaServer ES40, a Hitachi SR8000 and the Earth Simulator can not communicate with calculations. We propose a novel parallel strategy to perform the communication and calculations simultaneously by assigning the communication task into one of the physical processors installed on each node. Consequently, we confirm that the present method hides the communication cost behind the calculation one on a Compaq AlphaServer, a Hitachi SR8000 and the Earth Simulator.
We propose the Mixture Multi-component Extension (MME), which is an extension of the Lattice Boltzmann Method (LBM), to obtain a multi-component LBM model from a single component LBM model. There are advantages of MME that it has ability of coupling components with different masses, flexible application to various LBM models, and simulation of thermal miscible multi-component fluid. In addition, MME is able to solve molecule diffusion and convection of each component simultaneously.
This paper presents a computational scheme for compressible ideal MHD (magnetohydrodynamics) equations. The scheme utilizes the finite volume approach with an adapted grid refinement technique. The second order accuracy of the scheme in space and time is ensured by a MUSCL-type method and an explicit Runge-Kutta scheme for the time discretization. Computational results are presented for a MHD shock tube and two-dimensional supermagnetosonic flows.
In the present paper, numerical scheme for compressible ideal MHD equation using HGA Method are presented. In recent numerical simulations, efficiency of grid generation and efficient use of limited computer resources are demanded. In the present paper, the kinds of cell, which have great effect on analysis accurate and efficiency, are paid attention to. Therefore in order to take advantages of both shape flexibility and computational efficiency Hybrid Grid which is composed of Triangle, which takes advantage of shape flexibility, and Quadrangle, which takes advantage of computational efficiency, are used. Moreover in order to calculate more efficiently, HGA Method included AMR (Adaptive Mesh Refinement) Method is used, which obtains higher resolutions by using fine grids only in the location where the change of the physical quantity is intense, based on Hybrid Grid.
This paper is intended as an investigation of the implicit computational algorithm and the consistent (algorithmic) tangent modulus for finite strain single crystal plasticity. Although numerous studies have been made on an implicit computation algorithms, there is still room for improvement in a time integration of the rate of plastic deformation gradient. We here provide a new return-mapping stress update algorithm and a consistent tangent moduli in conjunction with the direct use of exponential representation. We carry out the numerical experiments on a single or polycrystalline metal and demonstrate that our proposed algorithm provides better approximation and does not bring much difference in the computational efficiency in comparison with the conventional approaches.
In this paper, the effect of the shape and distribution of anisotropic ellipsoidal inhomogeneity on the overall elastic moduli of a composite material is studied by equivalent model using the concept of local region from the viewpoint of micromechanics. The elastic moduli of such a material are theoretically analyzed, focusing on the cases of anisotropic ellipsoidal inhomogeneities or isotropic spherical inhomogeneitiesrandomly distributed in an isotropic matrix. Furthermore, using the example of SiC fibers or spherical SiC randomly distributed in an isotropic aluminummatrix, numerical results are provided. As a result of this study, it is found that the elastic moduli of such a material can be determined for an arbitrary volume fraction and shape of ellipsoidal inhomogeneities.
Particle methods are meshless methods in which motion of continua is simulated with a finite number of particles. In this paper, we develop a particle method from a direct discretization of the Lagrangian for inviscid incompressible flows. This particle method is a Hamiltonian system with holonomic constraints which stem from the incompressibility condition. In the accompanying paper, the RATTLE algorithm which is a symplectic scheme for constrained Hamiltonian systems is adopted to improve the quality of the particle method.
In the preceding paper, momentum equations of a Hamiltonian particle method are derived from direct discretization of the Lagrangian for inviscid incompressible fluid flows. They are a Hamiltonian system with holonomic constraints. In this paper, the RATTLE algorithm which is a symplectic scheme for constrained Hamiltonian systems is adopted to improve the quality of the particle method. Then traveling surface waves are simulated to examine the Hamiltonian particle method. The result shows the superior mechanical energy conservation property of the method.
Diagonally relaxed RIC factorization which revises relaxation on amount of modification for diagonal entries to improve convergence rate is effective preconditioning for CG method for solving linear system of equations with symmetric positive definite coefficient matrix. This preconditioning, however, demands much memory due to fill-in appeared during factorization. In this paper, we verify performance of RIC(0) factorization with no fill-in and modified RIC(0) factorization using acceleration parameter. Numerical experiments show that the proposed modification is extremely effective for gaining high convergence rate of the CG method.
This paper presents hybrid computing approach to solve inverse problem. An application in Electrical Impedance Tomography (EIT) with Newton's method is employed to introduce the new approach. With hybrid computing approach, accurate Jacobian matrix in Newton iteration can be obtained through derivative operation on explicit expressions between a particular response and a set of system parameters in EIT, which is impossible to be achieved by conventional numerical computing approach. Comparing with symbolic computing and general hybrid computing, limited variables are kept symbolic in order to decrease computing complexity. The simulation results have illustrated that the hybrid computing approach proposed in this paper is valid for solving inverse problem of Electrical Impedance Tomography.
One of coauthors stated that in case of framed structures under static loading conditions, the summation of the product of response sensitivity coefficients and sensitivity variables over all of the structures equals response values multiplied by a constant, and that constant value differs depending on the selected sensitivity variables and the order of sensitivity coefficients. We call the property which sensitivity coefficients have, the characteristic of sensitivity coefficients. The characteristic of sensitivity coefficients stated above, exists also in case of sensitivity coefficients of eigenvalues and eigenvectors. In this paper these characteristic of sensitivity coefficients are summarized. And the physical meaning and applications of the characteristic of sensitivity coefficients are explained.
We have developed a parallel visualization system that allows the user to optimize the parallel processing easily according to the data size and the system environment. This system can be serially processed to small data, can be processed in parallel to big data, and can be processed in batch to huge data which exceeds the total amount of memories. Moreover, in order to carry out parallel optimization according to the characteristic of visualization, the arbitrary parts of visualization processing can be parallelized. As a result of examining using this system, it turned out to big data that sort-last parallel rendering is effective. 26GB data exceeding system capacity was visualized by batch processing divided into 8 times.
Numerical methods for Finite Element Method (FEM) analysis of the exterior Helmholtz problem with local and Dirichlet-to-Neumann (DtN) boundary conditions are reviewed. Efficient preconditionings of Conjugate Orthogonal Conjugate Gradient (COCG) method for the resulting complex-symmetric (non-Hermitian) matrix systems are discussed including Incomplete Cholesky (IC) and Complex Shifted IC factorization with dropping tolerance. We evaluate the convergence of preconditioned COCG methods under varying wave numbers of Helmholtz equation and consider also on effectiveness of these preconditionings through eigenvalue analysis.
In recent years, global warming becomes a big problem and the reduction of carbon dioxide (CO2), which is typical of greenhouse gases, is required. Therefore two reduction methods are proposed. Namely, Type 1: Liquid CO2 is covered with CO2 hydrate, and is stored at the bottom of the deep sea. Type 2: CO2 is dissolved in seawater and is diffused in the sea. However, a simulator that can predict the behavior of CO2 in the current accurately does not exist. This paper proposes an agent-oriented simulator for estimating the behavior of CO2 in the current and the detail of the system is described. CO2, pipe, sea that are the essentials of this simulation are defined as the agent, respectively. This simulator has been developed by Java language and common agent classes are provided by us, and the operation has been confirmed through test cases. The result obtained from the test cases shows that the proposed simulator is suitable for estimating behavior of CO2.
Mesh superposition method has been only used for static analysis. In this paper, we provide the formulation of modal analysis using mesh superposition method. Additionally, we examine whether the local mesh can add a local shape like a hole to the global mesh. As a result, natural frequency corresponding to the domain cut off by a local mesh was obtained with that of main model. We provide the filtering technique to extract only the natural frequency of a main model using eigenvectors.
In this study, we formulated the Dynamic SGS model for Large Eddy Simulation in MHD turbulence at high magnetic Reynolds number using the GSMAC (Generalized Simplified Marker and Cell Method)-FEM. The FEM scheme was verified through the isotropic MHD turbulence. In most cases, the MHD turbulence was analized with the Spectral method using spectral cut off filter. So, few reference data for the MHD turbulence exist using Gaussian filter. In this study, we constructed the reference data in case of using the Gaussian filter and the data was used for the verification of the LES scheme based on the GSMAC-FEM. Through the verification, we examine the validity of the numerical scheme based on the GSMAC-FEM for MHD turbulence at high magnetic Reynolds number.
An Edge tone is self-excited sound, which is one of the important subsonic aerodynamic noises. This Edge tone problem phenomenon is computationally simulated using the two-dimensional compressible Navier-Stokes equations. The distance from the slit to the edge is changed with the jet speed and the slit width being kept constant. Acoustic waves having dipole characteristics are well simulated which shows edge tone mechanism is easily captured. FFT analysis shows that the frequency jumps agrees with the Brown's empirical formula. The frequency of the edge tone is influenced by the distance from the slit to the edge. The result indicates that the numerical approach presented is useful and become a good research tool for understanding if the acoustics problem.
An iterative method, namely COCG (Conjugate Orthogonal Conjugate Gradient) method, is applied to finite element sound field analysis of rooms. First of all, convergence of COCG method applied to finite element sound field analysis of three rooms is investigated from the following viewpoints: (1) sound source frequency; (2) absorption coefficient of a wall; and (3) sound source location. The results show that: (1) A lower frequency engenders faster convergence of COCG method in all rooms. (2) A higher absorption coefficient engenders faster convergence of COCG method except for a small cubic cavity. (3) Convergence of COCG method is affected by the sound source location in rectangular rooms. As an application of the results, we propose efficient parallel computation method using the relationship between frequency and convergence of COCG method. In addition, the results obtained by the numerical experiments for the three rooms are validated on a large-scale room.
The Bi-Conjugate Gradient method or Lanczos algorithm is an important guideline of a generalized CG method. In particular, BiCGStab method is among well known methods for solving a linear system of equations with unsymmetric coefficient matrix. However, the occurrence of breakdown of the methods can cause failure to converge to the solution. In this paper we present algorithm of new product-type BiCG method based on the minimization of an associate residual vector in place of the residual vector. Numerical experiments show safe convergence of the proposed BiCGSafe method as compared with the existing GPBi-CG and BiCGStab methods.
A computer-assisted liaison system is discussed and proposed in this paper in a distributed problem solving environment (PSE) for partial differential equation (PDE) based problems. In our previous paper (Trans. JSCES No. 20010018, (2001)) we have proposed a concept of distributed problem solving environment, which consists of several modules. The modules are distributed on network-linked computers and are registered on the distributed-PSE server so that available modules are shown to users. The distributed PDEbased PSE inputs problem information, and outputs a program flow, a program source code and a document. In this problem solving process users find required PSE modules and design workflows to complete their problem solving tasks. In the distributed PSEs each module may be developed independently by engineers or scientists or users, and old modules are also re-usable. One of key issues in the distributed PSEs is connectivity among the distributed modules. In this paper we propose and construct a liaison system to connect modules. The computer-assisted module liaison system generates an adapter module among the distributed PSE modules. The adapter module generated by the module liaison system inputs output data from preceding modules and/or external modules, and connect the data to the input data for the next module. This module liaison system extends the potential capability of PSEs extraordinarily.
Exhaustively searching and analyzing bio-data are indispensable in researches on genome. In this paper we adopt grid technology for processing the large-scale and diversified bio-data, and construct problem solving environment (PSE) based on the grid for genome analysis system, which enables acquiring new results and comprehensive knowledge by linking bio-data based on workflow and web technology. Basic local alignment search tool (BLAST), which consumes a lot of computational time in genome analysis, is executed using distributed and parallel processing under the grid environment in this system. As a result, BLAST throughput time is improved significantly. Workflow system in this system is applied to designing a polymerase chain reaction (PCR) primer for single nucleotide polymorphism (SNP) typing. By applying this problem solving environment, productivity of the PCR primer design is improved remarkably.
The study of periodic structures having electromagnetic functions has undergone rapid development recently, and both theoretical and practical achievements are garnering increasing attention. These electromagnetic structures, called metamaterials, exhibit photonic and electric bandgap phenomena, where intrusion of electromagnetic waves of certain frequencies into a periodic structure is inhibited. Wave bandgap phenomena can also be observed in elastic waves. Topology optimization has been successfully utilized for the design of such bandgap structures, based on the concept of eigen-frequency optimization inside a predefined basic lattice. This study proposes a novel topology optimization method for periodic microstructures of electromagnetic materials using the concept of propagation behavior to formulate design problems. The proposed method is also capable of automatically generating 2D periodic photonic bandgap structures without relying on a predefined lattice, and physically reasonable results were obtained.