Although the research about a legged type robot’s ZMP (Zero Moment Point) had many examples of research in the past, in those researches, main subjects are that the solution could not be really obtained 1 by 1. It is because redundancy is in the solution by the model of a multi-link system like a legged type robot. In this paper, formulas RHC(Receding Horizon Control) equal constraints are added, and it proposes about the method of controlling ZMP on real time. The proposed technique can generate the reference trajectory of ZMP on real time. As for formulization, both of simulation results are described a 2-dimensional plane and in the 3-dimensional space. This technique is easily biped, quadruped, and the legged robot of other types applicable.

抄録全体を表示

In this paper, material non-linear analysis considering tensile fracture was developed from the viewpoint of hybrid displacement model. In this model, the concept of the spring of RBSM was introduced in Lagragian multiplier. In addition, compatibility of the displacement on the intersection boundary was approximately introduced using the penalty as a spring constant. First, the equivalent solution with the analytical solution was obtained when it examined the accuracy of elasticity solution using this model. As a result of comparison with the FE analysis, this model had the accuracy equal to constant strain element of finite element method. Next, as a result of the slip analysis, the slip lines and the collapse load were equal to RBSM. Finally, the equivalent tensile collapsing load with FEM was obtained when it carried out analysis of the bending problem for the plain concrete beam.

抄録全体を表示

A wavelet-based preconditioning technique for conjugate gradient Poisson solvers is described. While the linear systems can be solved with an iterative matrix solver, the convergence speed becomes worse and the computing time increases with increase of grid points. The problem, however, can be solved by our preconditioning method using wavelets. In this paper, we specially pay attention to the application of Haar wavelets. Since the discrete wavelet transform has a good property in data locality, our algorithm can take advantage of parallel processing capabilities and enhance parallel performance, unlike many other preconditioning methods which are not suitable for vector and parallel processing. These kinds of combined approaches of software and hardware are superior in large-scale problems and will play more important roles in future computational science.

抄録全体を表示

A Lagrangian numerical simulation technique to calculate concentration fluctuation fields by a non-buoyant plume in grid generated turbulence has been developed. Lagrangian velocities of the fluid particles in the downstream decaying turbulence are simulated by the summation of unsteady random Fourier modes. The distribution of wavenumbers are selected to estimate the effects of large scale structure well. Then, the backward diffusion technique of particle-pair is applied to the prediction of scalar statistics. Simulations of scalar fluctuation fields of a line source plume in a grid-generated turbulence have been made and the simulation results are compared with the experimental data. the present new model gives good predictions of the actual scalar fluctuation statistics.

抄録全体を表示

In the previous papers[2,3], relationships between the stress rates and the tangent stiffness’ with the special orthogonal group SO(3) for finite element formulation were derived. These stress rates were Truesdell stress rate, Jaumann stress rate, Neo-Green stress rate and linear stress rate. And Euler-buckling problems[1] by beam elements using these stress rates were analyzed and the buckling loads were compared with the theoretical solutions[4]. Then, in order to compare with those theoretical solutions[4], theoretical buckling loads using these stress rates are obtained in this research. At first, tangent stiffness’ of beams using four kinds of stress rates are obtained in this paper. Next, equilibrium equations of these beams will be obtained and these equations of Euler-buckling loads will be obtained in the next paper[5].

抄録全体を表示

In the previous papers[2,3], relationships between the stress rates and the tangent stiffness’ with the special orthogonal group SO(3) for finite element formulation were derived. These stress rates were Truesdell stress rate, Jaumann stress rate, Neo-Green stress rate and linear stress rate. And Euler-buckling problems[1] by beam elements using these stress rates were analyzed and the buckling loads were compared with the theoretical solutions[4]. Then, in order to compare with those theoretical solutions[4], theoretical buckling loads using these stress rates are obtained in this research. At first, tangent stiffness’ of beams using four kinds of stress rates are obtained in last paper[5]. Next, equilibrium equations of these beams will be obtained and these equations of Euler-buckling loads will be obtained in the this paper.

抄録全体を表示

The structural identification and dynamic control of a cable-stayed bridge are considered to be difficult due to the structural complicacy and system uncertainties. In this paper, based on the concept of decentralized information structures, a decentralized non-parametric identification and control method is proposed with neural networks for the purpose of suppressing the vibration of a well-studied and documented two-cable-stayed bridge induced by earthquake excitations. The control strategy proposed here uses the stay cables as active tendons to provide control forces through appropriate actuators. Each individual actuator is controlled by a decentralized neuro-controller that only uses local velocity and relative displacement information. The feature of decentralized control simplifies the implementation of the control algorithms and makes decentralized control easy to practice and cost effective. The effectiveness of the decentralized identification and control algorithm is evaluated through numerical simulations. The performance of the decentralized control is compared with that of the linear decentralized control method. And the adaptability of the decentralized neuro-controllers for different kinds of earthquake excitations is demonstrated via simulations.

抄録全体を表示

In these days personal computers (PCs) became an essential part of today’s school education. However teachers or lecturers meet difficulties to obtain the information about how all the students operate computers or which softwares they use. In this work we construct a teaching assistant system which acquires information about the student’s computer operation and analyzes the information. The system, that is called a TSUNA-TASTE, consists of four major components. The first component is a distributed agent that is installed in each student’s personal computer and acquires the student operation information. The second component is a teaching assistant server that communicates with the agents and analyzes the information. The third component is a database system which stores the student operation information and other configuration data. The fourth is a communication system using a WWW server and a browser. In this paper we present and discuss a design concept of the system and the details of each component.

抄録全体を表示

The ladder sleeper is a newly developed rail-support component for the next generation railway structure. The floating ladder track using the ladder sleeper proposed by authors is floated from the concrete track-bed by vibration isolators, that are devised instead of ballast to reduce total weight and the vibration of the track. A drop-weight impact test machine is fabricated on the floating ladder track to study on the impact behavior of the track. The experimental results are compared with finite element results. An efficient finite element formulation and the numerical method to solve the combined dynamic behavior of a Shinkansen train running on the track with rail irregularities in railway structure are presented. Based on the numerical method impulsive wheel-load characteristics caused by the high-speed train running on the track with rail irregularities are investigated.

抄録全体を表示

Composite shells have excellent structural performance and there is a growing tendency for structural designers to adopt more composite structural systems in Japan. However, many types of these kinds of structures have many problems concerning their fire resistance. It is important to establish the valid methods for rational evaluation of fire resistance of these structures. In this paper, computational methods to evaluate the fire resistance of concrete structures are presented and some computational examples are shown. The purposes of this study are; 1) to present the method of heat and moisture transfer analysis, 2) to propose the elasto-plastic themal stress analysis method taking into account of the moisture effects. The validity of these analysis methods for concrete members under high temperature is evaluated.

抄録全体を表示

This paper presents a numerical analysis of incompressible viscous fluid flow using higher-degree interpolation scheme. This scheme is considered to be one of the generalizations of the IDO (interpolation differential operator) method. The solution procedure is composed of the third order upwinding scheme for advection term and the high order scheme for non-advection term in governing differential equations. Effectiveness of the method is shown with numerical solutions of the cavity flow problem for Re=1000,5000 and 10000 in comparison with the results given by Ghia et al. And we show numerical solutions of flow past two prisms for another numerical example.

抄録全体を表示

This paper describes a local contact search method using a multilayer neural network. Contact search process consists of two phases: a global search phase for finding the nearest node-segment pair and a local search phase for finding an exact local coordinate of the contact point within the segment. In the present method, the multilayer neural network is utilized for finding the local coordinate of the contact point in the local search phase. The fundamental formulation of the neuro-based local contact search method is described in detail, and its basic performance is demonstrated through some analyses.

抄録全体を表示

Previous works on fuzzy regression models have been developed as an extension of the interval regression model. On account of this line of approach, in previous works, observed data of dependent variables are not contained in minimizing objectives but in constraint conditions only. As a result, the estimated fuzzy numbers that are derived from those models can be said that they have little reflection of the dependent fuzzy numbers used for identifying models. In this paper, the authors propose a method for identifying fuzzy regression model. Our method is based on the concept of least square estimate as well as interval linear regression of the previous type, and includes two types of problems to solve. The first type problem is given as a single-stage problem, and the second is given as a two-stage one. Since observed data are explicitly taken into objective functions in both problems, fuzzy regression models identified by our method are more sensitive, with respect to observed data, than those of previous method.

抄録全体を表示

This paper describes a parallel explicit finite element analysis system for dynamic problems based on the hierarchical domain decomposition. Here a whole mesh of analysis domain is first decomposed into several large domains called Parts, each of which is again decomposed into a number of smaller subdomains. The proposed parallel finite element analysis is performed on three layer processors : one GrandController, several Controllers and a number of Analyzers. Finite element analyses of subdomains are performed on the Analyzers, then the analysis results of the inter-subdomain DOFs are collected to the Controllers, and finally those of inter-Part DOFs are collected to the GrandController. The present method shows good scalability because of its reduced and distributed inter-processor communications. Parallel performances are tested through elastic wave propagation analyses on two typical parallel processing environments, i.e. a PC cluster and a massively parallel computer.

抄録全体を表示

Recently, the thermal therapy using HIFU (High Intensity Focused Ultrasound) is noticed as an effective therapy for cancer in the medical field. In this therapy, it is important to examine the transmission of the ultrasound and the temperature distribution of the tissue around the cancer. The computational simulation is the best tool for such examination, because it does not injure the human body. However, the calculation cost becomes enormous in the general-purpose solver, because the wave length of ultrasound is very short. Therefore, in this study, we develop an effective simulation solver using the voxel finite element method. In this method, the calculation cost is as low as the finite deference method, and the robustness of the finite element method is combined. In this paper, the FE formulation of wave propagation analysis of ultrasound with perfectly matched layer (PML) absorbing boundary is presented. The effectiveness of the present method is demonstrated by some numerical examples.

抄録全体を表示

The overlaying mesh method is an effective method for the modeling of the design. In this method, a local model is overlaid on a global model, the former represents high resolution of deformation and the latter represents rough deformation. In this paper, we theoretically prove that if we make a hole only on the local model, the stress of the hole on the global model becomes zero. And then, the result becomes equivalent to the result in which the hole exists actually.

抄録全体を表示

An iterative domain decomposition method is applied to numerical analysis of time-harmonic 3-dimensional eddy current problems, which is related to complex symmetric linear systems. The iterative domain decomposition method is combined with the BiConjugate Gradient (BiCG) procedure, and the Hierarchical Domain Decomposition Method (HDDM), which has been shown effective for structural and heat conductive problems, is adopted for the parallel computing. Numerical results show that the BiCG procedure converges and that a transformer model whose number of complex degrees of freedom is about 100 thousand can be solved efficiently by using HDDM.

抄録全体を表示

In this paper, we present and construct a distributed problem solving environment (PSE) for partial differential equation (PDE) based problems. The system inputs a problem information including a discretization and computation scheme, and outputs a program flow and also a C-language source code for the problem. On a host computer a user inputs his/her problem, and the host navigates the user to solve the problem. The distributed PSE for PDEs consists of several modules: a problem description module, a discretization one, an equation manipulation one, a program design one and a program generation one. Each module is distributed on distributed computers, and the whole information is described by the Extensible Markup Language (XML) including the Mathematical Markup Language (MathML); each distributed module communicates with the host module by using XML documents, so that outputs from each modules are visualized. Independent modules which are developed by other engineers or users for one of the functions specified above can be also used after adjustments to the distributed PSE interface, if necessary. Therefore the concept of the distributed PSE extends the potential of conventional PSE systems. The system is implemented by the Java language with the Document Object Model (DOM), in order to make the system a networked and distributed PSE.

抄録全体を表示

In this paper, we develop new discrete method which applied the approach of the hybrid model to a seepage flow problem with phreatic surface. First, the formulation which introduced the potential continuity by penalty function in general weighting residual procedure is shown. Compatibility of the potential on the element boundary edge is approximately introduced using the penalty function. In this method, an independent potential field is assumed for each element, and the head and the hydraulic gradient of each element are used as the degrees of freedom. Therefore, the shape of the element is not limited and the node does not need to have the degree of freedom and is used only for recognition of the domain shape. Then, we developed the technique in which only the integral region changed without initial mesh division in order to decide the phreatic surface. The equivalent solution with the analytical solution is obtained when it examined the accuracy of numerical solution using this technique.

抄録全体を表示

Fully developed three-dimensional turbulent flow over sand ribbons in an open-channel is computed to clarify the generation mechanism of secondary currents of the second kind by direct numerical simulation (DNS) using a regular grid under a generalized curvilinear coordinate system. Stable secondary currents and turbulent characteristics are reasonably reproduced. Detailed evaluation of each term in a mean vorticity transport equation revealed that the secondary currents of the second kind is generated principally by cross-planar turbulent shear stress and that the anisotropy of the cross-planar turbulent normal stresses restrains the generation of the secondary currents.

抄録全体を表示

On designing and developing a transformer, the finite element analysis of three dimensional (3-D) eddy current problems is very useful. This paper shows a large scale 3-D eddy current computation for a transformer tank. Results show good agreement with those of the previous computation, which considered the 1/6 part of the whole transformer.

抄録全体を表示

AMG (Algebraic Multigrid) method is one of the most promising solvers for large-scaled unstructured problems. In this study, the performance of AMG solver has been compared with that of the classical single-grid solver used in GeoFEM; PCG (Preconditioned Conjugate Gradient). The benchmark problems are 2-Dimensional (2-D) steady heat conduction, 3-D homogeneous or heterogeneous penetrative flow and several 3-D elastic problems of FEM. As the result, though AMG requires more memory compare with PCG, AMG is faster than PCG for all large cases which have large degrees of freedom.

抄録全体を表示

A visualization system VIS has been developed to visualize simulation results of a Shinkansen train running on the railway structure at high speed. The VIS based on an object-oriented description of the structure includes Shinkansen train, the railway structure, and the surrounding environment to visualize. Visualization of a Shinkansen train running on a cable stayed long-spanned concrete bridge in Nagano Shinkansen line is demonstrated.

抄録全体を表示

This paper describes the parallel finite elements for ultra large scale analysis for a elasticity problem using massively parallel computers. In order to solve the issue of memory shortage and computational time, the developed system employs an estimated static load balancing technique and hierarchical distributed data management. The present system is successfully applied to large FEM analyses of over 100 million d.o.f. (degrees of freedom) with high parallel performance over 80% on a massively parallel computer consisting of 1,024 processing units.

抄録全体を表示

For analyzing huge, complex phenomena, the loosely coupling analysis method on parallel computer is very effective. In this paper, a program and a library which support to develop such kind of coupled program are investigated, described and evaluated. Especially, because the spatial relation analysis among meshes is thought as a very time consuming process, the detailed performance analysis and improvement of such a process have been done. The program is developed as a part of GeoFEM, which is intended to become the software platform for the solid earth problems on ES40.

抄録全体を表示

The mechanical behaviors around a crack tip for a system including both a crack and two tilt grain boundaries under uniaxial tension are obtained using a molecular dynamics simulation. The Johnson potential for α-Fe is used in this analysis. Three kinds of the grain boundaries are examined. The <110>{112} grain boundary has the lowest grain boundary energy among all tilt grain boundaries. The <110>{111} grain boundary is one of the grain boundaries which have local-maximum grain boundary energy. The <110>{332} grain boundary has grain boundary energy between the preceding two grain boundaries. In a system including both the crack and the <110>{112} grain boundaries, not only a phase transition from bcc to hcp but also ductile deformation occurs around the crack tip in order to relax stress concentration. Dislocations emitted from the crack tip are observed and two dislocation pile-ups near the grain boundaries are formed. Moreover the phase transition domain at the crack tip changes to twin deformation after the elapse of time. In a system including both the crack and the <110>{111} grain boundaries, a new crack is generated at the grain boundaries and intergranular fracture occurs. In a system including both the crack and the <110>{332} grain boundaries, dislocations are emitted not only from the crack tip but also from contact points of a free surface and the grain boundary, then complex dislocation structure is formed.

抄録全体を表示

We have developed and designed a problem solving enviroment(PSE), called DISTRAN which generates a fortran simulation program based on partial differential equations(PDEs). One of key issues in PSE and also scientific simulation is to determine a stable time-step size Δ t and to reduce computation CPU time. Usually Δ t is common for one set of grids. In this paper we propose an Individual Time Step(ITS) for Runge-Kutta Method, in which each grid has an individual Δ t which is variable depending on each grid point. In DISTRAN, each individual-time-step Δ t is automatically determined to be stable in Runge-Kutta Fehlberg Method. Even if PDEs have stiff problems, we can solve the difficulty of stability and reduce the CPU time in DISTRAN. DISTRAN implements the ITS method automatically and provides a fortran program to users.

抄録全体を表示

This paper is concerned with the structural damage detection of five-story frame model by free vibration analysis, in combination with algorithm based on filtering theory. The structural damage analysis is generally formulated as an inverse problem. The formulated inverse problem must be analyzed under the consideration of stochastic properties on the response of structural system. In this study, three kinds of filtering algorithm based on the extended Kalman filter, projection filter and parametric projection filter are employed as the analytical algorithm of the inverse problem. Characteristics of those three kinds of analytical algorithm are discussed from results of the inverse analysis on the five-story frame model.

抄録全体を表示

At recently, electronic machine systems etc are requested more complicated and complex construction for the manufacture of more higher functional system. However, the previous development technology cannot created the high functional system than in this time. Therefore, we focused the generative complexity of living organisms for the obtain the process procedure of higher complexity. In order to elucidate the generation mechanism of biotic complex system. We proposed the simulation of bionic generative complexity based on Immunologial self recognition and evolutional process in this paper.

抄録全体を表示

Configuration and control of magnetic fields in a magnetron sputtering system are studied by the Finite Element Method (FEM) caluculations. The configuration of magnetic fields is found to be essential to confine electrons. The modification of magnetic field configuration is studied to enhance the generation of plasmas. The problems of plasma generation and erosion are studied by tracing an electron in a magnetron sputtering system with a ferromagnetic target.

抄録全体を表示

A numerical method to simulate the flow induced by the Marangoni convection and the surface tension with the free surface has been developed using the GSMAC-FEM. Level Set Method was applied to capture the free surface and CIP-FEM was used for the advection equation governing a kinematic condition of the free surface, re-initialization for the distance function and the total volume conservation. Oscillation of a liquid droplet was calculated and highly accurate volume conservation was performed. Numerical analysis of the thermocapillary flows in a square cavity was performed whether the surface tension term is treated as the body force or the surface force. Both results were good agreement with each other. Numerical simulation for Marangoni convection in a liquid bridge between two discs was successfully simulated.

抄録全体を表示

In case that only nodes on the surface of a 3-dimensional analysis domain are given as input data for the mesh generation, how to generate its surface becomes very important for finite element users. The authors propose two procedures to cover the surface by triangles using only nodes on the surface of the analysis domain. At the beginning the authors explain the idea of the surface generation from nodes, and successively describe the methods of surface generation in details. Main difference between two methods is that one of them requires some additional information in case of boundary configuration with geometric complexity. Both methods are based on Delaunay triangulation, which accepts only nodal coordinates and divides the convex domain covered by nodes into tetrahedra geometrically. Therefore, the success of the surface generation wholly depends on the placement of nodes on the surface, and the paper includes how the nodes should be placed on the surface.

抄録全体を表示

In this paper, a norm scaling preconditioning algorithm for the conjugate gradient method is proposed in the finite element analysis. Although the present scaling algorithm requires the same memory size and computational time in each iteration step as the usual diagonal scaling preconditioning algorithm, the number of iterations to convergence of the former is smaller than that of the latter. The implementation of the method on parallel environment and some numerical examples are also described to verify the effectiveness of this preconditioning algorithm.

抄録全体を表示

We have examined, in a previous research, how the quality of a finite-element solution depends on the method of connection from a fine mesh region to another region to be twice as coarse, in a case of plane strain and a use of the four-node quadrilateral element. In this paper, above discussion is extended to a case using degenerated solid element. Three processing manners of the transition region of the mesh density are discussed. Namely, 1) a primitive meshing which filled up the quadrilateral elements with trapezoidal shape in the transition region; 2) a jointing manner such as a large element and two small elements are connected directly with compatible condition of inter-element displacement; 3) special elements having five-node are arranged for the region. Numerical examples show that a method using a five-node element having piecewise linear shape function and a primitive method provide good results, and direct joint method gives poor.

抄録全体を表示

We present a three bodies trading model in markets for a numerical simulation of financial markets. It has three different type agents who correspond to daytraders, market-makers, and investors. Market dynamics can be explained as the result from market participants acting. The model analyzed all trades by non-statistical and non-stochastic approach: calculating interactions among three agents. Practically we develop a computer scheme with combining the presented model and generic algorithms[1]. We show the scheme is able to simulate major stocks in Tokyo Stock Exchange accurately. The scheme can simulate not only the price movement, but also the volume series simultaneously.

抄録全体を表示

A medial surface generation technique for generating a shell model for finite element analysis of a thin structure has been developed. This method has the advantage of reducing analytic work load. It consists of four steps; first, a pair of thin faces is found; second, a generatrix medial surface of the pair is generated; third, an intersection line between the generatrix medial surface and boundary face of the solid is calculated; fourth, an intersection line is trimmed off the generatrix medial surface. The effectiveness of the technique was confirmed by applying it to the medial surface generation of magnetic disk head parts.

抄録全体を表示

In many mechanical structures, structural design for noise reduction is becoming increasingly important. Noise reduction is often achieved through structural modifications. However, it is hard to predict the effectiveness of noise reduction by typical approaches. This paper presents an optimal design approach for reducing sound radiation from a vibrating panel by its curvature design. We use the calculation procedure which consists of three parts, the shape representation technique using a spline function, the coupled structural-acoustic analysis based on FEM and an optimization technique based on a genetic algorithm (GA) which are performed in an iterative loop. The integrated program developed herein is utilized for the analysis of booming noise in truck cab which is caused by resonance of the engine compartment. It is shown that the curvature design of the floor panel obtained by using this method can achieve effective reductions in the sound pressure level at the receiver point.

抄録全体を表示

This paper presents an optimization method of production efficiency and the difference of the production time of each machine (line balance) about the placement time in the chip mounter system. Placement time changes greatly by the performance of the chip mounter and the configuration of the system. The problem is how to distribute parts and parts feeders to each chip mounter. We propose a distributing method using GA which makes it attain two purposes. Several numerical experiments about some systems composed of more than one chip mounter with different performance were done. It confirmed that our proposed method had usefulness more than conventional method through the results of numerical experiment. Furthermore, those characteristics and usefulness are shown about the GA adopting simple constraint method (SCM-GA) and the GA adopting improved constraint method (ICM-GA) that a constraint is changed one after another with the progress of the generation.

抄録全体を表示

Earth simulator is a distributed memory parallel system which consists of 640 processor nodes connected by a full crossbar network. Each processor node is a shared memory system which is composed of eight vector processors. The total peak performance and main memory capacity are 40Tflops and 10TB, respectively. A performance prediction system GS³ for the Earth Simulator has been developed to estimate sustained performance of programs. To validate accuracy of vector performance prediction by the GS³, the processing times for three groups of kernel loops estimated by the GS³ are compared with the ones measured on SX-4. It is found that the absolute relative errors of the processing time are 0.89%, 1.42% and 6.81% in average for three groups. The sustained performance of three groups on a processor of the Earth Simulator have been estimated by the GS³ and those performance are 5.94Gflops, 3.76Gflops and 2.17Gflops in average.

抄録全体を表示

The tangent stiffness method provides a useful and simple algorithm to find shapes of curved surfaces, which can be applied to membrane and cable net structures. This paper proposes a control technique of loading and displacing, when the method is applied to shape analyses under inner pressure. The proposed technique, named “Simultaneous control”, can find high-rise and large volume shapes, which are in the state over the maximum point of its p-v pass. The technique controls both of compulsory displacements on a control point and the inner pressure derived from the shape in each iteration step at the same time. As a result, when the unbalanced forces on every node converge as well as the average of converted inner pressure, the rational form is found.

抄録全体を表示

We developed a new design system that enables designing high-performance pumps in a shorter period of time. This system contains a digital engineering tool, which automatically constructs a new 3D-CAD model of the pump impeller after evaluating the flow path shape. The tool was connected to the 3D-CAD/CAE coupling system and used to reduce the time of the design process. The reliability and effectiveness of this system was tested by using it to design a general-purpose small pump. We found that the analytical work time decreased to 2/5 at the first operation and to 1/6 after the second operation. The system was also connected to optimization software and tested in terms of its contribution to improving pump efficiency with regard to other requirements.

抄録全体を表示

This paper describes a new neural network modeling technique, named the Self-Designing Neural Network, which incorporates an integrated learning algorithm for structure and weight parameters. Neural networks are frequently applied to various problems such as optimization, control, prediction and data mining because of their nonlinear mapping capability. Although neural networks have wide application, in a multi-layer neural network the establishment of structure parameters such as the number of layers, units and connections between units depends on the experience of the engineer and hence the accuracy of mapping is reduced in many cases. The proposed neural network modeling technique, the concept of which is based on ATR’s neurite network, was developed in order to avoid reliance on engineers. In the proposed method, each neuron grows an axon and a dendrite, and each connection between neurons is automatically constructed in the emergent process. The structure and weight parameters of the neural network in each problem are determined by optimizing the growth process using a genetic algorithm. In this study, the optimization is realized by parallel computation. The validity of the proposed method are investigated by applying the method to Exclusive OR.

抄録全体を表示

Object Oriented Programming (OOP) has been proposed for developing sophisticated software that needs future expansion. Successful applications of OOP in different fields such as graphic data processing and basic OS etc. is intensively reported in the literature. To apply OOP technique in softwares for structural analysis, the target structure is represented by the basic objects (beam, truss, etc.) using key concepts such as encapsulation, polymorphism and inheritance. These features facilitate easy development and management of the program. Yagawa and Kanto⁽¹⁾ proposed class structure in structural analysis program to apply OOP technique. They showed advantages of OOP for a two-dimensional linear problem. Referring their work, we have developed seismic design program for bridge, in which some new classes, class members, class member functions were developed. It has been shown that with newly developed classes, class members and class member functions and OOP techniques, the modification and expansion of program can be more easily achieved comparing with other researchers proposal, and of course traditional procedural programming technique.

抄録全体を表示

The structural damage detection of truss structures is formulated as the framework of inverse problem. In this inverse analysis, the solution procedure is developed in combination with the forward structural analysis and the filtering algorithms. Especially, the filtering algorithms are based on the filter theory of the well-known Wiener filter and the projection filter. Paying attention to some notable characteristics in the truss structure in which the change of stiffness under damage of member is lager than that of the frame structure, the natural frequencies corresponding to each eigen-mode are effectively used as the observed data. In order to verify the characteristics of both filters two types of truss structure are selected in performing numerical example. The first one is a symmetric truss whose both ends are simply supported. And the other is a cantilever-type un-symmetric truss. The notable characteristics of two filtering procedures in applying this inverse problem are made clear through numerical simulations.

抄録全体を表示

We develop a method of topology optimization for structures at finite strains based on the mathematical homogenization theory. After formulating the geometrically nonlinear problems with hyperelastic material, we briefly review the conventional approach that employs orthotropic homogenized properties. We then propose to use finite number of sampling points for updating design variables in a finite element by introducing isotropic homogenized material properties, whereas the conventional approach uses one set of design variables in an element. The corresponding formulation approach is consistent with the mechanical problem modeled by the homogenization method. Several representative numerical examples are presented to show the validity and applicability of the proposed approach. In particular, we try to clarify the mechanism which generates the optimal structure prevented from revealing structural instabilities such as buckling.

抄録全体を表示

In a rural town, there is an obvious decline in power as people moves away from the central area. Although autonomy is measuring the recovery of power by the law such as many promotion law as the correspondence to such a present condition, it needs to grasp the present condition of the town as a fixed quantity in order to correctly measure the effect.<br> The purpose of this report at this time is to evaluate historic old houses and tree group from the position of calculation engineering and to show the expression of Self-Organized Criticality with the aim of actual application to (among other fields) social engineering. Basically the model seeks to apply the Hausdorff s-dimensional measure (adjusted using the Power-low expression) to both historical old houses and tree group separately and then add the corresponding figures and calculate the answer by the Biplot analysis, which is the principal component of a low dimension analysis. The Hausdorff s-dimensional measure is then combined with the low dimension measure to provide a logarithmic expression, which will be a measure of the presence of Self-Organized Criticality. This can then be compared with results from the virtual town block, which produced its results using pseudo random numbers.

抄録全体を表示

We suggest expansion methodology for Black-Scholes equation in financial engineering field with hydrodynamics interpretation. Black-Scholes equation is famous for derivative pricing model, introducing from Ito’s process in theory of probability, and it is applying in financial area widely. Because of Black-Scholes equation based on ideal assumption, financial engineers have been devising parameters or have been creating parameter structure in the equation in order to explain financial market dynamics. They mainly have focused on diffusion term in stochastic process that is recognized as volatility on their studies. We apply the concept of advection term for financial market, and expand Black-Scholes equation. Obtaining equation can express market dynamics simply and explain derivative market distortion, volatility smile.

抄録全体を表示

We showed new financial market simulation methodology and its application result. We developed the three agents model that was suggested in previous paper[1], and enhanced intraday simulation scheme in order to discuss market complex structure. Although there are many approaches to analyze financial market, we indicate agent simulation is one of the most powerful tools to elucidate financial market structure, particularly in intraday movement study. Applying new scheme for Japanese equity market, intraday fluctuation can be simulated qualitatively. Lastly we discuss existence of multi scenarios in financial market and refer to the possibility of multi scenario structure.

抄録全体を表示

In this paper, we propose a Dirichlet Neumann domain decomposition finite element/boundary element coupling method with initial Dirichlet data assumed on the FEM/BEM interface approached from the FEM sub-domain. The method accounts for situations where the Neumann boundary conditions are specified on the entire external boundary of the FEM sub-domain. We also investigate the convergence of the method. The theoretical analysis provides an interval from which the relaxation parameter has to be chosen in order to achieve convergence.

抄録全体を表示