The particle method (SPH and MPS method) shows the good performance to simulate the drastic fragmentation of water. While, the surface tension has not been taken into consideration in the most previous studies on the particle method. Hence, in some previous simulations, it was difficult to reduce effectively the small scale disturbances on water surface including a small fragmentation, although the occurrence frequency of which was not so large. In this viewpoint, the accuracy of the water surface tracking in the scale of particle diameter was not necessarily satisfactory in the previous simulations by the MPS method. In this study, the surface tension model, which is the body-force-type description of the surface tension proposed by Nomura et al., is modified to fit the calculation of the wave breaking and runup process in hydraulic engineering.

View full abstract

This paper reports current status of parallel computer simulations for superconducting state dynamics on the Earth Simulator. The simulated subjects are divided into three themes as follows, 1) preliminary simulations for neutron detecting devices using MgB_2 superconductor, 2) fundamental simulations for studying effects of thermal fluctuation and magnetic field on the stability of superconductivity, and 3) test simulations for novel devices proposed by making matrices composed of various superconductors with different symmetries. The equations employed commonly in the three themes are the time-dependent Ginzburg-Landau equation, the Maxwell equation, and the thermal diffusion equation, and the matrix diagonalization is also performed to study full quantum features for the themes 2) and 3). Now, we have developed two codes for solving the three equations and excuting the matrix diagonalization, respectively. On the Earth Simulator, the former code show a performance exceeding 18Tela-flops at the maximum, while the latter code succeeded to obtain an eigenvalue of a matrix with 180 million dimension.

View full abstract

This paper presents large-scale parallel simulation of seismic wave propagation in Japan using the Earth Simulator. The seismic wavefield in 3D heterogeneous structure is calculated by solving equation of motions in elastic media by using the parallel FDM. The results of application of the parallel FDM for modeling strong ground motion from recent large earthquakes such as for the 2000 Tottori-ken seibu earthquake in Japan, and comparison between high-density seismic observations (K-NET and KiK-net), demonstrating the effectiveness of the simulation model. Therefore the model can be applied to estimate the pattern of strong ground motion damage for future earthquake scenarios.

View full abstract

This paper describes our parallel visualization system for large unstructured datasets on the Earth Simulator. Parallel volume rendering is one of important methods in the system. Some strategies have been adopted to make it suitable for the Earth Simulator including concurrent visualization with computation for extremely large data size, supervoxel-based partition for complicated unstructured grids, detail-preserving resampling to improve visualization quality, and optimization techniques according to the hardware features of the Earth Simulator. Good visualization images and high speedup performance on the Earth Simulator demonstrated the feasibility and effectiveness of the proposed methods.

View full abstract

NEXST_BCtool is a multipurpose boundary condition attachment tool. We developed this tool to attach boundary conditions to large scale analysis model which has hundreds of millions degree of freedom. Boundary contition input interface is written by solver developer with simple text format and users can input boundary condtiions with unified interface for several analysis problems and solvers. We use Java language and Java3D to port for several OSs and window systems easily and get comfortable hadling response.

View full abstract

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, the material micro property such as crystalloid was neglected. In order to obtain valuable result, another method, in place of the conventional finite-element method, must be considered. As an altemative method, molecular dynamics method is significatly implemented in the field of micro-cutting, indentation and crack propagation. Moreover, since speed of computer came to considerable fast and parallel computing technology grew mature in the last two decades, tendency of CAE (Computer Analysis Engineering) toward micro level or macro level become stronger. In this paper, the phenomenon of ultra-precision grinding with sub-nanometer chip thickness is studied on basis of molecular dynamics. In addition, general parallel computing technology, spatial decomposition is endeavored to improve the calculation efficiency.

View full abstract

[in Japanese]

View full abstract

In the finite element method errors occur depending on the number of elements and their arrangement. It is very important for optimal design to estimate these errors. For this reason, a posteriori error estimation methods have been developed. In structural problems many of these error estimation methods are for static problems. However there are few error estimation methods for eigenvalue problems. In this paper a method of a posteriori error estimation for eigenvalue problems is presented. This method is based on the fact that eigenvalue problems are equivalent to the minimization of the Rayleigh quotient. This estimation is done by element-wise static analysis and correction using a convergence order of an a priori estimate. Simple numerical examples are shown.

View full abstract

Recently, the sound and viblation analysis on computers are ubiquitous. The degree of freedom of figure model for analysis are increasing year by year. By the performance of the present analysis technique and capability of a computer, It is difficult to sound as actual sound from the result of analysis. It is unsatisfactorily for referring to the result of an experiment and obtaining. However, desiners are looking forward to reprocudt the sound from analysis results and they expects the computer will be stronger more than current. So, we have to prepare how to design the sound of structures. In this research, We process making a hole, a rib, a patch, bending, on figure model. We search how it change the low frecquancy part of the sound and the middle frequency of the sound.

View full abstract

There is a big issue to generate 3D hexahedral FE model, since a process to divide the whole domain into several simple-shaped sub-domains is required before generating a continuous mesh with mapped mesh generators. In general, it is nearly impossible to set up proper division numbers interactively to keep mesh connectivity between sub-domains on a complicated arbitrary-shaped domain. If mesh continuity between sub-domains is not required in an analysis, this complicated process can be omitted. Usually penalty scheme is applied to set up constraints to keep the continuity between two discontinuous individually meshed sub-domains with different division numbers, however this approach results in an unacceptable rigid deformation in most cases, due to the characteristic on the system of constraint equations. FFGM might accept discontinuous meshes, which only requires nodal information. However it is difficult to choose a reasonable influenced domain in moving least squares scheme with non-uniformly distributed nodes in discontinuous FE models in reality. The blending shape function with refined numerical integration was originally proposed in two dimensional QUAD element by Gupta in 1978,but this approach is originally limited only to 1-iiregular mesh type in three dimensional HEXA element as the work by Morton in 1995. The extension of this function is a promising way, because there are no parameter dependencies. To apply blending shape function to arbitrary HEXA elements in the discontinuous parts of mapped mesh above, element-wise local surface meshing is needed to construct serendipity blending HEXA element. Here a new finite element shape function for discontinuous mapped HEXA8 mesh with element-wise surface-mesh based serendipity hierarchical blending function is proposed to solve the problem mentioned above.

View full abstract

In this paper we illuminate problems of using statistics in mesh enhancement process and compare two possible adaptations. The case of 2-dimensional triangular mesh where aspect ratio value is chosen as mesh quality criteria parameter is discussed. Due to considered methods the opportunity for existence of proper geometrical realization for given mesh topology is judged.

View full abstract

There are some difficulties in Digital Design. First, in initial design period when CAD data are not yet enough, CAD/CAE system is not applicable for the design. Second, it is very difficult to transfer between different type CAD systems. Here we try to use CG based on CSRBF (Compactly Supported Radial Basis Functions) for efficient digital design.

View full abstract

The project "in silico Hip Joint" aims to construct a model of "each patient" in a virtual space (in silico) and to make Surgical Planning of Hip Joint. First, by analyzing the 3D model constructed from CT images, several parameters for the surgical operation are measured and the techniques of doctors are considered objectively. Second, Finite-Element models are constructed for the consideration from mechanical viewpoints. Lastly, the feasibility of the remote surgical planning is tested.

View full abstract