Compliant mechanisms are a new type of mechanism, designed to be flexible in order to achieve a specified motion as a mechanism. This paper presents a new structural optimization method for the design of compliant mechanisms based on the level set method. First, the basic details of a structural optimization method based on the level set method are briefly discussed. Next, an optimization problem that addresses the design of compliant mechanism considering the ratio of the displacement at the input location to the displacement at output point, and topological derivatives for this design problem are formulated. Based on this formulation and the level set method, a new structural optimization algorithm is constructed where the solving of the equilibrium equations and the updating of the level set function are performed using the FEM. Finally, several design examples are provided to confirm the usefulness of the proposed structural optimization method.

View full abstract

This paper presents a simple FEM-based analysis method for crack propagation analyses of heterogeneous solids. In order to develop a robust and simple algorithm for handling multiple crack growth and interfacial debonding involved in the fracture analysis of heterogeneous solids such as concrete materials, we dare to stay with the standard FEM incorporated with the nodal-integration and the mesh re-alignment techniques. After presenting the problem formulation together with the penalty method, we detail the computational implementations for representing the propagation of strong discontinuities. Benchmark problems, characterized by mixed-mode fracture, are solved to validate the proposed method and numerical examples demonstrate that the present procedure is robust, stable and effective even for the fracture analysis of two-phase composites including multiple cracks and interfacial debonding.

View full abstract

The present study involves a novel numerical technique regarding the simultaneous use of the fast multipole method (FMM) and a special-purpose computer originally designed for molecular dynamics simulations (MDGRAPE-3). In the present calculations, the dynamics of two colliding vortex rings have been studied using the vortex method and the computation time has been reduced by a factor of 2000 compared to a direct calculation on a standard PC. The reconnection of the vortex rings was clearly observed, and the discretization error became nearly negligible for the calculation using 10⁷ elements.

View full abstract

It is well-known that Conjugate Gradient (CG) method with DIC factorization can be viewed as one of useful solvers for a linear system of equations with a symmetric positive definite matrix. On the other hand, Eisenstat’s trick also can be viewed as an efficient implementation for CG methed without preconditioning only. However, this technique had limited effectiveness because of insufficiency and instability of DIC factorization. In this paper, we will propose a strategy for a more efficient and stable implementation combined with accelerated DIC factorization and Eisenstat’s trick and by numerical experiments we will present that this indeed may lead to faster and stable convergence.

View full abstract

The fluid-shell coupled overlaying ALE mesh method proposed by the authors is developed by introducing the force distribution technique of the immersed boundary (IB) and immersed finite element (IFE) methods to couple interface-tracking local mesh and non-interface-fitted global mesh without projections of fluid velocities and pressures. The proposed method can be understood as an improved version of not only the single ALE mesh method but also the IB and IFE methods that are different fluid-structure simulation technique. Computational performance of these methods is compared in a two-dimensional bench mark problem with a large deformation, and it is shown that the proposed method has superior performances to other methods from several aspects. An essential factor to the stable analysis by non-interface-fitted approaches are also investigated in the paper.

View full abstract

In this paper, we present an improvement of wall boundary calculation model for Moving Particle Semi-Implicit (MPS) method. Although particle methods are well suited for computation of a free surface flow, they still have issues. Wall boundaries are usually descritized into particles and used as boundary conditions. To compute a fixed boundary, the particles are calculated as fixed particles. These wall particles cause several issues. First, the wall particles, which are a descretized expression of boundary, cannot represent a smooth boundary. When a boundary with a slope is considered, they are not calculated as a smooth slope but a bummpy surface. Second, the wall particles increase the computational burden and so this leads to increase of computation time because the number of wall particles are large. The method proposed in this paper does not use wall particles but uses polygons. To compute each term, the contribution of fluid particles and the boundary is separated. This paper shows how the equations are derived from usual equations and then shows some computational results with the present method. Lastly, the result using the present method is compared with one using the wall particles quantitatively.

View full abstract

Compliant mechanisms are a new type of mechanism, designed to be flexible to achieve a specified motion as a mechanism. Such mechanisms can function as compliant thermal actuators in Micro-Electro Mechanical Systems (MEMS) by intentionally designing configurations that exploit thermal expansion effects in elastic material when appropriate portions of the mechanism structure are heated. This paper presents a new structural optimization method for the design of compliant thermal actuators that are constructed based on the level set method and the Finite Element Method (FEM). First, an optimization problem is formulated that addresses the design of compliant thermal actuators considering the magnitude of the displacement at the output location. Next, the topological derivatives that are used when introducing holes during the optimization process are derived. Based on the formulation and the level set method, a new structural optimization algorithm is constructed that employs the FEM when solving the equilibrium equations and updating the level set function. The re-initialization of the level set function is performed using a newly developed geometry-based re-initialization scheme. Finally, several design examples are provided to confirm the usefulness of the proposed structural optimization method.

View full abstract

A unified algorithm for compressible and incompressible flows is developed for Moving Particle Semi-implicit (MPS) method. Original MPS method was used for incompressible flows. MPS has been used to analyze complex flows, such as ship motion in ocean waves, droplet breakup and other complicate flows involving free surfaces, fluid breakup and coalescence. We extend MPS to manage compressible flows as well as incompressible flows in the same way as C-CUP (Cubic Interpolated Pseudo particle Combined Unified Procedure) method. We term it MPS-AS (MPS for All Speed) method. In MPS-AS method, compressible mass, momentum, energy conservation and gas state equations are used as governing equations. A Poisson equation is derived, using an evolution equation of pressure. This equation includes terms related to sound speed. These terms become prominent when a time step is small compared to sound speed. On the other hand, the discretized equation approaches MAC (Marker And Cell) method as the time step gets larger. The present algorithm can be applied to both incompressible and compressible flows including shock waves. Two-dimensional dam break and shock tube problems are solved. As a result, MPS-AS method can capture free surface (incompressible flow, dam break) and expansion wave, contact face and shock wave (compressible flow, shock tube).

View full abstract

In this paper, we propose a new setup tool for a PC cluster system called DCAST. DCAST can setup a PC cluster system in an intermingled environment of diskless and diskfull nodes. The unique function of DCAST is the mechanism to design the network of diskfull nodes and diskless nodes. The network is automatically optimized to avoid loads from concentrating on a single diskfull node when the diskless nodes are booted. To verify the utility of this function, we measured the file read access speed and the processing time of applications for cluster, and verified fault-tolerance of the diskless nodes. The results showed that the file read access speed improved by approximately 30%, the processing time improved by approximately 5%, and the number of system failures decreased when the concentration of loads to a single diskfull node was avoided. Therefore, DCAST is useful for constructing large-scale PC clusters.

View full abstract

The Generalized Minimal RESidual (GMRES) method with variable preconditioning is an efficient method for solving a large sparse linear system Ax = b. It has been clarified by some numerical experiments that the Successive Over-Relaxation (SOR) method is more effective than Krylov subspace methods such as GMRES and ILU(0) preconditioned GMRES for performing variable preconditioning. However, SOR cannot be applied for performing variable preconditioning, when solving the linear system with zero diagonal entries. Therefore, we propose to make good use of a strategy proposed by Duff and Koster, namely, an algorithm for permuting large nonzero entries onto the diagonal in order to enable SOR to be used for performing variable preconditioning. By numerical experiments, we show the efficiency of the variable preconditioned GMRES using SOR when applying the algorithm for permuting large nonzero entries onto the diagonal.

View full abstract

This paper presents a new time-domain boundary element method based on the Operational Quadrature Method (OQM) in 2-D antiplane viscoelastodynamics. The conventional time-domain BEM approach cannot be used in general, since it is difficult to obtain the closed time-domain fundamental solutions in viscoelastic wave propagation. To overcome the difficulty, in this paper, the Operational Quadrature Method (OQM) developed by Lubich is applied to 2-D antiplane viscoelastodynamics. In the proposed method, the convolution integral is numerically approximated by quadrature formulas, whose weights are computed by using the Laplace transform of the fundamental solution in 2-D viscoelastodynamics. Moreover, the Fast Multipole Method (FMM) is adapted to improve the computational efficiency for large scale problems. As numerical examples, viscoelastic wave scattering by many cavities is shown to validate the proposed method.

View full abstract

Wind turbine blades made of CF/GF hybrid blades are expected for the next generation large wind turbine. The hybrid blade is made from Carbon Fiber Reinforced Plastic and Glass Fiber Reinforced Plastic, and structural optimization of the blades is a multi-objective optimization problem (MOOP) on weight reduction and cost reduction. In the present paper, the MOOP is solved using Multi-Objective Genetic Algorithm (MOGA). The turbine blade structure has to satisfy constraints with respect to ultimate strength, fatigue strength, buckling and deflection. To evaluate these constraints, full scale FEM analyses are indispensable, and. this requires enormous computational cost. In this paper, to reduce the cost, Kriging model response surface approximations for the surrogate model of the constraint satisfaction are constructed. Since Pareto solutions obtained using Kriging model response surface do not always satisfy the constraints, the obtained results are confirmed to check the satisfaction of the constraints using FEM analyses. Using the FEM results, Kriging response surfaces are improved to fit more accurately around the Pareto optimal results. Some new approaches to judge for improvement of Kriging response surface on designing wind turbine blade are tried. As a result, CF/GF wind turbine blade is successfully optimized with low calculation cost.

View full abstract

A novel finite volume method has been presented to solve the shallow water equations on triangular unstructured grid. Keeping both PV and VIA as the model variables and updating the boundary PVs through a local Riemann solver make the formulation much easier to be extended to unstructured grids with local reconstruction over single grid cell. We present a two dimensional formulation for shallow water equations on triangular unstructured grid of high order accuracy. Numerical oscillation can be effectively eliminated through the use of the reconstructionsv of oscillation free interpolation functions.

View full abstract

A dental clinical support system to estimate the mechanical stress in mandibular bone with implants was manufactured in a manner readily accessible for practical application, as an integration of the existing techniques established in the area of computational engineering. A mandibular bone model was constructed from a set of X-ray CT images using a threshold-based binalizing, and finite element mesh was generated with the use of a 3D-CAD software and an FE mesh generator. The proposed system was capable of generating FE meshes of the mandibular bone reflecting the individual geometry of a patient and analyzing the mechanical stress around the dental implants exerted by an occlusal force. The finite element analysis system developed herein was applied to the estimation of the optimal alignment and design of implants. This system has thus turned out to be useful in the evaluation of mechanical properties of a mandibular bone with implants.

View full abstract

MPS (Moving Particle Semi-implicit) method is one of the particle methods, which can be used to analyze incompressible free surface flow without surface tracking. However, MPS causes unreal pressure oscillation with high frequencies. We propose a new formulation for the source term of the pressure Poisson equation. The source term in the original MPS method is devided into three terms. With proper relaxations of the three terms, we can suppress the pressure oscillation. Smoother pressure distributions are obtained in a hydrostatic pressure and dam break problems.

View full abstract

In this paper, an optimization for particle-based simulations on the Graphics Processing Unit (GPU) is presented. The sliced grid is a good candidate to make the search for the neighboring particles efficient because it improves the memory efficiency of the uniform grid and also improves the performance. However the previous study used graphics functions such as alpha blending, it is not clear the sliced grid is also suited for general streaming processors. So we present a general implementation of the sliced grid and an implementation using Compute Unified Device Architecture (CUDA). This paper also proposes the block transition sort that is well suited for GPU utilizing the coherency between simulation timesteps. The block transition sort is used to improve the memory alignment of the simulation data, i.e., to increase the spatial locality of the data. Distinct Element Method (DEM) is implemented to evaluate the proposed methods. We achieved about 3x speed up for the neighboring particle search which is the most expensive part of the particle-based simulation and about 1.5x speed up for the overall computation.

View full abstract

A new characteristic/Lagrange Galerkin finite element method using an Hermitian type element, named Hermitian characteristic Galerkin (HCG) method, is proposed for solving advection diffusion problems. Using the Hermitian type element can improve accuracy of the characteristic Galerkin finite element method. A composed function for each element included in the formulation of the characteristic Galerkin finite element method can approximate a polynomial function by constructing the Hermitian type element in upwind position, and it is possible to perform integration of the term included in the composed function analytically. The 10-DOF Hermitian type triangular element which consists of cubic polynomials given by function values and first order derivatives on each vertex and a function value on barycenter of a triangle is used in the two dimensional calculation. Two numerical experiments for the two dimensional advection diffusion equation are investigated and the results show the effectiveness of the present method by evaluating the computational accuracy.

View full abstract

We have conducted research and development of the Cognitive methodology based Data Analysis System (CDAS) which supports researchers to analyze large scale data efficiently and comprehensively. In data analysis, it is important to evaluate whether data is valid or not and whether data is meaningful from a scientific viewpoint. Traditionally, much of analysis process has been carried out by humans. However, when the scale of data is extremely large, data analysis is beyond the recognition capability of humans. The basic idea of CDAS is that computers execute data analysis instead of humans. In the present study, we analyzed data analysis process and showed three necessary functions (Verification and Validation: VV, Data Diagnosis: DD and Synthesis) for the design of CDAS. VV and DD functions are executed on gird computing environment and findings output from these functions are integrated on Synthesis functions. Finally CDAS displayed only useful information to human. We have applied the system to the virtual plant vibration simulator and succeeded in analyzing large scale data reaching to 1TB thoroughly for the first time.

View full abstract

Plug row model, which is a dynamical model for coal and gas outbursts, is developed to locally opened system. At the first report, the author lead the model with the assumption of the locally closed system. But for the case of the mass of gas in a local system changes, the system is no longer the closed system. At that condition, the model has to be developed to the locally opened system. There are two typical cases that the system has to be treated as the locally opened system. One is gas flow between local systems, the other is gas desorption from a coal. These effects are added to our model. With new parameters for gas flow and gas desorption, numerical simulation is examined. And in-door experimental results, which is performed by another researchers, is compared with our numerical calculation results.

View full abstract

In this research, a new technique of the particle method is proposed for the reduction of communication between nodes in the PC clusters. Particle numbers are rearranged along one axis using a bucket. This reduces the amount of communication between nodes. The maximum and minimum particle numbers in the communication list are memorized. A simulation of fluid dynamics using 6,300,000 particles was carried out as an example, and it was shown that the efficient parallel calculation was possible in the large-scale analysis.

View full abstract

This paper presents a new time-domain boundary element method based on the Operational Quadrature Method (OQM) in 2-D viscoelastic wave propagation. The conventional time-domain BEM approach cannot be used in general, since it is difficult to obtain the closed time-domain fundamental solutions in viscoelastic wave propagation. To overcome the difficulty, in this paper, the Operational Quadrature Method (OQM) developed by Lubich is applied to 2-D viscoelastodynamics. In the proposed method, the convolution integral is numerically approximated by quadrature formulas, whose weights are computed by using the Laplace transform of the fundamental solution in 2-D viscoelastodynamics. Moreover, the Fast Multipole Method (FMM) is adapted to improve the computational efficiency for large scale problems. As numerical examples, viscoelastic wave scattering by many cavities is shown to validate the proposedmethod.

View full abstract

Edgetone phenomenon is computationally simulated using the two-dimensional compressible Navier-Stokes equations with high-order compact finite difference scheme and Runge-Kutta time integration scheme. Flow features obtained in the present computations agree well with that obtained in the experiment in the past. The results show that the edgetone frequency decreases with jet Mach number. This indicates that compressibility effect exists in the edgetone phenomenon, which justifies the feedback-loop mechanism proposed in the past. Detailed analysis of convective velocities of the jet disturbance is then conducted. There exists non-linearity; the convective velocities are high near the jet exit and low near edge. The result indicates that the value 0.25 proposed by Powell as a phase-lag parameter based on the assumption of constant disturbance velocity includes this non-linear effect, and the value of the real phase-lag parameter becomes -0.2 when considering non-linearity.

View full abstract

Recently P. Sonneveld and M. van Gijzen proposed new IDR theorem extended from the original theorem. Their technical report, however, has imperfection of the IDR theorem and lack of description on implementation. In this paper, we consider two types of preconditioing of IDR(s) method. One of preconditionings is refered to as an easy preconditioning, and the other is refered to as a profound preconditioning. We consider on reliability of two types of preconditioning in view of accuracy of the approximate solutions and computational cost. In particular, we make clear that the amount of computational cost of right preconditioning is the least amount compared with left and two-sided preconditionings. Through two kinds of numerical experiments, we verify that profound right-preconditioning of IDR(s) method outperforms among other iterative methods from the viewpoint of CPU times and amount of memory.

View full abstract

A method of parameter identification by means of numerical material test data is developed for anisotropic hyperelastic materials within the framework of the homogenization method for heterogeneous solids with periodic microstructures. The method is a standard least-square scheme, but is based on the tensor representation of the macroscopic nominal stress and the macroscopic deformation gradient calculated from numerical material tests. The specific error function is proposed for approximating the anisotropic hyperelastic constitutive equation which is linear in its material parameters, whereas a set of loading patterns relevant to parameter identification is determined on the analogy of the method for homogenized elasticity constants in linear elasticity. Several numerical investigations are performed to validate the proposed method and demonstrate its promise and potential.

View full abstract

In this paper, a method to stabilize and smooth pressure distribution in Moving Particle Semi-implicit method is proposed. Incompressible condition is represented by coupling of two traditional conditions. Divergence-Free condition is used to calculate pressure smoothly and particle number density condition is used to keep the volume constant. Quasi-compressibility is considered to make the calculation more stable. A dam break is simulated more stably and pressure distribution is calculated smoothly. Surface particles are detected more accurately. Moreover, the calculation cost is lower than the traditional method.

View full abstract

We developed a new method for the analysis of the dynamic systems of elastic materials which include backlashes by using Moving Particle Semi-implicit (MPS) method. Rigid-elastic interaction model is newly introduced for the effective analysis. We applied the method for the analysis of the concrete cask subjected to seismic motions. The seismic data which was observed in Kobe in 1995 is used in the present study. The rocking angle and displacement of the concrete cask is evaluated by the simulation and compared with those in literature. It is shown that our model can perform the simulations effectively and the results are in good agreement with those of the reference data.

View full abstract

Distributed Network Application System (DNAS) is a P2P-oriented middleware which supports to obtain system information and to exchange data among several applications on Grid. DNAS constructs a dynamic communication topology for fault tolerance using tree topology and provides API for acquisition of system information and application data. Users can develop an application that adapts to a dynamic change of the calculation resource by using DNAS. In this paper, we proposed the design and implementation of DNAS and developed a max number search application in the DNAS environment. Moreover, we evaluated the basic performance of DNAS functions.

View full abstract

It is one of the important research subjects to solve large deformation FSI (Fluid-Structure Interaction) problems, especially those where a large-deformable thin structure moves in a highspeed flow field, e.g., airbag deployment. In this study, A fixed fluid mesh with the level set function is applied in order to analyze the FSI problems. The level set function is one of the implicit function and can express the large-deformable interface as zero isosurface. A partitioned solution method (iterative staggered scheme) based on fixed mesh with the level set function is constructed in order to combine advanced fluid and structure solvers, i.e. CIP-FEM for fluid and structural elements considering large displacement/rotation increments for structure. In the present method, it is important to update the level set function according as the deformable interfacial geometry and deal with kinematical condition at the interface. To address this issue, we propose an interface treatment using level set virtual particles. The particles are arranged in the normal direction to the interface. It is verified that the present method has enough accuracy and stability for the large deformation FSI problem.

View full abstract

This paper presents a multi-scale analysis method for thermo-elastic coupled problems of porous media based on the homogenization method, which accounts for the effects of the size of periodic microstructures or unit cells. The size effect under consideration is supposed to be caused by the amount of microscopic heat transfer that depends on the micro-scale pore size of the porous media. We first formulate the multi-scale multi-physics problem between heat diffusion and deformation of porous media based on the asymptotic homogenization method in consideration of the size of unit cells with pores on the surface of which heat is transferred. Then the macroscopic heat conduction phenomena are simulated with unit cell models of different sizes to demonstrate the promise and potential of the proposed method in comparison with the corresponding single-scale direct analyses with detailed numerical models.

View full abstract

The effects of boundary conditions of computational analysis on edgetone phenomena are investigated. Two-dimensional compressible Navier-Stokes equations are solved with high-order compact finite difference scheme and Runge-Kutta scheme. The edgetone phenomena under various nozzle lip-thicknesses and jet profiles are computed. Edgetone phenomenon becomes stronger with increasing nozzle lip-thickness. This effect seems to be explained as larger lip-thickness emphasizes feedback-loop. Besides the change of jet-profile under the same maximum velocity affects edgetone frequency, strength and stage. These effects are due to the difference of averaged velocity. The change of jet-profile under the same averaged velocity affects frequency of the edgetone. This effect is caused by the difference of convective velocity of disturbance with Powell's feedback-loop theory.

View full abstract

As the present computing capacity has continued to grow, it has been possible to apply numerical simulation for various phenomena such as fluid dynamics and solid dynamics. Because the failure analysis of structures must deal with dynamic phenomena which involve large deformation, the treatment of plastic mechanics is important. In this research, new method for 3-dimensional elastic-plastic analysis is developed based on Moving Particle Semi-implicit (MPS) method. The strain is modeled using gradient model of the MPS model and the stress is evaluated based on Von-Mises theory using elastic predictor-radial corrector method. The new method is applied for two cases. It is shown that the results have a good accuray.

View full abstract

We apply a newly developed characteristic-curve finite element scheme to cavity flow problems. The scheme is useful for large scale computation, because P1/P1 element is employed and the matrix of resulting linear system is symmetric. Numerical results of two- and three-dimensional cavity flow problems are presented. Three types of the Dirichlet boundary condition, discontinuous, C⁰ and C¹ continuous ones, are treated, and the difference of the solutions is discussed.

View full abstract