This paper considers parallelism of AINV (Approximate INVerse) preconditioning by means of grouping of factorized matrices. The AINV preconditioning consists of independent each other factorized matrice. Therefore, the AINV preconditioning can parallelize easily than the conventional preconditionings, e.g., ILU(0) and SSOR precoditionings. Through numerical experiments, we demonstrate high efficiency of the proposed AINV preconditioning for parallel computation.

View full abstract

  We consider on reduction of computational cost for solving both real and complex linear systems. In general, the cost of complex operations needs several times larger than that of real operations. Therefore, we multiply the imaginary unit of i to the original linear systems. Through numerical experiments, we examine that the proposed technique of implementation is very effective in order to solve efficiently both real and complex linear systems.

View full abstract

  Recently the numerical simulation plays an important part in the emergency management for sediment-related disaster. Water and sediment go down streams that have conjunctions, curves, narrow and wide parts, so that the numerical simulation has to treat the complex geometry. Moreover, it is necessary for the numerical simulation to be applied for not only very fast flows such as debris flows and floods but also very slow flows such as still water. The numerical simulation needs to be based on unstructured mesh and to be improved in order to handle very slow flow such as still water. This study aims to develop a numerical model for still water and to verify the numerical model by reproducing the experimental results. In the developed numerical model, it is assumed that the averaged gradient of flow surface is equal to the resultant force of the hydrostatic force acting on the boundaries of element. It is found that the numerical simulation with the developed numerical model can reproduce the still water and the surges triggered by the flow going into the reservoir.

View full abstract

  We consider on reduction of computation time utilizing homogenous compiler and execution on other computers. We treat with matrix-vector multiplication used always in the algorithms of many iterative methods, e.g., Krylov subspace methods. In general, computaion of matrix-vector multiplication needs much time. Through numerical experiments, we make clear that the strategy of utilization of homogenous compiler and execution on other computers is effective.

View full abstract

  Recently, miniaturization, low cost and standardization activates of the wireless sensor network technologies has been remarkably advanced, because of advancement of the wireless technologies. Especially smart grid is important application of the wireless sensor network technologies. In the smart grid network, the network congestion occurs caused by huge traffic in real time transmission of energy usage data or environment information data in the house or building. It is required to effectively collect the data and improve the network efficiency. We describes the scheme and the results in this paper since we clarified the efficient data transmission based on data fusion method for the energy usage data using Back Propagation (BP) algorithm with Artificial Neural Network.

View full abstract

  The requirements for wireless sensor networks are low power consumption and long lifetime because sensor nodes are not always supplied by mains electric power. To reduce power consumption, sensor nodes have to reduce the idle listening duration. In this paper, we propose a MAC protocol without using carrier sensing. Packets of all sensor nodes are randomly transmitted within a frame which length is as long as data collecting interval. We perform the theoretical analysis of our protocol and derive the successful transmission rate in case that packets are transmitted in different duration. We simulate our protocol to verify the theoretical analysis.

View full abstract

  We have investigated the sedimentation phenomena of a colloidal dispersion composed of spherical ferromagnetic particles using Brownian dynamics simulations. We have attempted to clarify the conditions under which all particles subside and the thin film is formed, by varying the applied magnetic field, magnetic force between particles, mass density, and temperature of solution. The main results obtained are summarized as follows. As the mass density of the particles increases, the influence of gravity increases; hence, sedimentation velocity increases, which makes thin film formation easier. For strong magnetic force between particles, thin film formation is difficult, since the particles combined with each other to form clusters. However, if the applied magnetic field is strong and the magnetic moments of the particles are oriented in the direction of the magnetic field, the clusters disappear and thin film formation occurs, although the same magnetic force exists between the particles. Furthermore, the random force by Brownian motion is required for the cluster collapse and lead to thin film formation. However, if the influence of the random force is too large, the particles on the thin film will be unstable and there is a case where the particle moves upward.

View full abstract