This paper reports on the evaluation of deformation of the dissection cavity and changes in internal pressure in a chronic aortic dissection of DeBakey III type by applying fluid-structure interaction (FSI) analysis. FSI analysis is a numerical analysis method that includes the interaction between blood (fluid) and blood vessel (structure), and by solving the problem, it is possible to obtain the deformation and stress of the vessel wall and the difference in blood pressure between the true lumen and false lumen, which cannot be obtained by conventional methods that target only the fluid.

In the integrated modeling of a building structure and the ground, the most difficult thing is to maintain the consistency due to the large difference in mesh size at the boundary between the structure and the ground. The author proposed a modeling method that combines two inconsistent meshes using s-version finite element mesh superposition method with the local mesh exceeding. In the s-version FEM with the local mesh exceeding, since both of the two meshes represent only a part of the analysis model, it is not clear which mesh should be set as the global mesh in the formulation. In this paper, we report the effect of global / local settings on the analysis results in s-version FEM with the local mesh exceeding. Global / local settings have no effect on analysis in the formulation. It was confirmed in the numerical example.

We propose a novel shape optimization for designing a multiscale structure with desired static deformation. The square error norm between actual and target displacements of the macrostructure is minimized as an objective function. The design variables are the shape variation field of the outer shape of the macrostructure, the interface shape of the macrostructure and the shapes of pores of the microstructures. In this study, the macrostructure is divided into some arbitrary domains, which have independent periodic microstructures. The homogenized elastic tensors are calculated using the asymptotic homogenization method, and apply to the correspondent domains of the macrostructure. The shape gradient functions with the state and the adjoint variables are derived for the shape variation of the macrostructure and the microstructures are introduced, and apply to the H1 gradient method to determine the optimum shapes. The proposed method is applied to a both ends fixed beam and it is confirmed that the objective function decrease to zero, or the desired static deformation can be achieved as expected while obtaining the clear and smooth boundaries.

The authors have developed finite element analysis (FEA) program based on explicit method including modified incremental technique in order to shorten CPU time in material non-linear problem with cracking. Also, introduction of the method can be expected to improve convergence and stability of the problem. In this paper, we demonstrate the computational algorithm of the FEA program and verify the validity through a few numerical examples. In addition, we show its accuracy and CPU time in comparison with numerical results obtained from existing FEA programs, which have been developed be authors.

This paper develops an accelerated projected gradient method for solving the minimization problem of the total potential energy of an elastic body consisting of a bi-modulus material. It is known that this optimization problem can be recast as a semidefinite programming problem, which can be solved with a primal-dual interior-point method. Compared with this existing approach, the proposed method does not use the second derivatives of functions, and are free from numerical solution of linear equations. Numerical experiments demonstrate that the proposed method outperforms the semidefinite programming approach.

The numerical method of the implicit stress integration for the subloading surface model with the tangential-inelastic strain rate and the isotropic hardening stagnation in addition to the isotropic hardening and the translations of the back-stress and the elastic-core is proposed in this article.

The existing assembly method for carbon fiber composite (CFRP) requires bolts, which increases the assembly cost and reduces the strength due to drilling. Therefore, fusion bonding technology using induction heating which is enable to low-cost bonding without using bolts is expected, but the uniformity of the heating temperature in the bonding range is extremely important to obtain the desired bonding strength with this method. However, it takes a lot of costs and time to find the appropriate coil specifications for uniform heating from experiments or tests. In this study, using electromagnetic field & heat conduction coupled analysis and Bayesian optimization which is a type of response surface methodology, attempted the efficient estimate the correlation between coil specifications and heating temperature distributions. A program developed by Kyushu Institute of Technology was used for the analysis, and a Python library, PHYSBO, was used for optimization. The multi-objective optimization method used in this study was able to find improve solutions with a small number of searches, but only the three Pareto solutions were obtained. We plan to validate this study results with actual induction heating experiments and with another algorithm in the future.

In recent years, computational fluid dynamics has become more familiar in the industrial field, but most of the approaches are based on the assumption of Newtonian fluids. Resin and cement, which are widely used as industrial materials, are non-Newtonian fluids with viscoelasticity. The flow of viscoelastic fluids has some unique phenomena such as the Weissenberg effect. Therefore, there are many cases in which the behavior of viscoelastic fluids cannot be accurately reproduced or predicted by assuming a Newtonian fluid. Thus, with the aim of accurately simulating the behaviors of viscoelastic fluids with free surfaces, the moving surface mesh-incorporated particle method has been extended to viscoelastic fluids based on the Oldroyd-B model. The extended method is verified using the viscoelastic Poiseuille flow, where approximately 2nd order convergence is confirmed. Furthermore, we succeeded in reproducing the Weissenberg effect, which is a phenomenon unique to viscoelastic fluids. The climbing heights, calculated with varying Weissenberg number, are found to agree well with the analytical solution.

For CAE simulations based on the finite element method, double-precision real numbers, which requires more energy than others, have been almost exclusively used. On the other hand, due to overwhelming demand for deep learning, low-precision arithmetic is now being introduced into general-purpose processors as accelerators. With SDGs (Sustainable Development Goals) in mind, it is apparent that we must save energy as possible as we can. By using low-precision real arithmetic for finite element analysis, we can effectively use the arithmetic unit equipped for deep learning and get advantages in terms of calculation speed and power consumption. In this study, a new energy-saving finite element analysis assisted by deep learning is proposed. In the proposed method, deep learning indicates where to use lower precision computation for saving energy. Here, for the first try, the effect of low precision numbers on analysis results is quantitatively investigated through sample analyses.

In order to realize a multi-gas detecting sensor, the change of adsorption behavior of CO and NH3 molecules on graphene under uniaxial-strain was analyzed by using first-principles calculation. The adsorption energies E_ad well as Bader charge transfer ∇Q were calculated to evaluate the adsorption behavior and determine optimal adsorption configuration. Under the application of uniaxial tensile strain, the free energy of both adsorption systems increased monotonically with strain, and there was individual critical strain at which the adsorption energy changed from negative to positive, in other words, desorption started to occur. These results clearly showed that the adsorption behavior on graphene can be controlled by strain for multi-gas molecule detection.

Ferrous materials are widely used for various industrial products, of which mechanical properties depend on the microstructures initially determined in solidification process. According to recent studies using in situ observation, a new class of phase transformation takes place following solidification from BCC-δ phase to FCC-γ phase. This is referred to as δ-γ massive-like transformation which is different from peritectic reaction expected from Fe-C binary phase diagram. The δ-γ massive-like transformation may leads to surface cracking in casting, because this is a solid-solid transformation without any liquid phase and thus the strain energy which originates from density difference between the two phases would not be relieved. Thus, elucidation of the reaction mechanism is needed for the control of microstructures. Our previous studies revealed that this transformation is governed by continuous nucleation of γ phase and it is also suggested that the formation of vacancies would play important roles in transformation. However, it is difficult to directly observe nuclei or vacancies by in-situ observation. Thus, in this study, using phase-field-modeling, we analyzed the roles of vacancies on transformation, and found that vacancies are introduced in the generated γ phase above thermal equilibrium concentration. It is also suggested that the excess vacancies in the γ phase are caused by both vacancy formation in γ phase and diffusion from δ phase.

Numerical implementation of inactive elements in balancing domain decomposition method is presented. The inactive elements cause floating degrees of freedom, which make the coefficient matrix of the linear system of equations singular. Regularization methods for matrices related to the Schur complement, the diagonal scaling preconditioner, balancing domain decomposition (BDD) method and the BDD with diagonal scaling (BDD-DIAG) method are derived. The present numerical implementation showed good convergence performance in a numerical test of metal additive manufacturing simulation.

Dynamic wetting behaviors of water on the surface of structures are of great interest in industry. In this study, a liquid film flow on an inclined plate is numerically simulated by using the MPS method, focusing on the liquid dripping around the edge of the plate. In order to simulate such dripping behavior, it is important to calculate surface tension forces with taking the contact angle into account stably and accurately, under large free-surface deformations. For this reason, the surface tension forces are calculated using Kondo’s potential model, which is known as a highly robust surface tension model. Numerical simulations of a liquid film flow are carried out with varying inclination angle and contact angle. As a result, it is found that the occurrence of liquid dripping predominantly depends on the contact angle. The same tendency is also observed in experimental measurements.

This paper describes about mutual recognition between JSME Senior Analyst and NAFEMS PSE, benefit of PSE certification for PSE certified Senior Analysts, and PSEs who are contributing to the development of advanced semiconductor manufacturing equipment. Mutual recognition between JSME Senior Analyst and NAFEMS PSE has started since 2014 and number of International Senior Analysts who have been certified as PSE is forty-seven. Number of PSEs who have been certified as Senior Analyst is two. Mutual recognition has been achieved. Results of a questionnaire survey of International Senior Analysts by JSME showed 65% of respondents felt benefit of PSE certification. And most of respondents thought benefit of PSE certification could increase if awareness of PSE would increase. ASML, semiconductor lithography equipment maker, introduced customized PSE scheme for all simulation users in the company to acquire proper usage of simulation for development of semiconductor lithography equipment for most advanced semiconductor chips.

When structures are repeatedly subjected to unexpected excessive loads such as earthquakes, unstable failure occurs after low cycle fatigue. Therefore, there is a need for a technique to predict the fracture behavior using finite element analysis (FEM). Previous studies suggest that the shape of the crack front edge of the analytical model influences the results. Purpose of this research is investigating the simulation method of elasto-plastic fatigue crack growth focused on crack front shape. For this purpose, elasto-plastic crack growth tests were carried out on several CT specimens, and each test was terminated at various cycles to obtain the crack front shape. The fatigue crack growth tests were simulated using an analytical model that reflected the observed crack front edge shape and a model that averaged the front edge shape. The crack growth timing was adjusted and the effect of the frequency of crack growth in the simulation was investigated. The results suggest that the effect of crack front edge shape is small. It is also found that the fluctuation of the maximum load became smaller when the crack propagated every cycle.

In the casting process of alloys, semi-solid region is deformed due to the solidification shrinkage and/or external force, which causes serious solidification defects such as porosity, cracking, and macrosegregation. Better understanding of semi-solid deformation is crucial for reducing solidification defects. In this study, the multi-phase-field lattice Boltzmann model and simulation method, which can simulate the growth of multiple dendrites with the liquid flow, solid motion, and collision of multiple solid grains, is constructed to simulate the semi-solid deformation.

CFRP laminated composite structures manufactured by Automated Fiber Placement (AFP) may have gap and overlap defects and these potentially affect the strength of the structures. This study aims the development of finite element analysis models, which can consider strength reduction due to such defects. This paper proposes the modeling method considering ply-thickness variation induced by gap defects and shows the numerical results obtained by damage propagation analyses for Open Hole Tensile Test of CFRP laminate with a gap defect.

In computational fluid dynamics using particle method, many approaches have been attempted to enhance the accuracy or stability of calculations. These approaches included the improvement of the free surface boundary conditions. However, smooth distribution of the particles on the interfaces has not been researched enough. Thus, in order to improve the particle distribution on the interfaces, a particle exchange technique is developed and adapted to the simulation of Rayleigh-Taylor instability in the present study. Four methods of particle exchange, which are combinations of different approaches of the region of exchange particles and the evaluation of exchange, are tested. In all simulation, the interface of fluid is smoothed, and as well as the development of interface agrees well with the result of the conventional MPS method.

Prediction and control of dendrite growth is crucial for material properties. Phase-field method is widely used for the prediction of dendrite growth. Although the phase-field simulation of dendrite growth needs interfacial properties such as solid-liquid interfacial energy, mobility and these anisotropies, there are no those values for almost all materials. In this study, we predict the interfacial properties of pure Al by data assimilation and phase-field simulations based on the observation data computed by molecular dynamics (MD) simulation. MD solidification simulations are conducted by LAMMPS with EAM potential of pure Al. The four interfacial properties of pure Al are predicted by the data assimilation with ensemble Kalman filter and multiple phase-field simulations.

Phase-field (PF) method is a powerful numerical model for predicting dendrite growth during solidification. Although some material properties are required to perform PF simulations, the lack of those properties, especially the interfacial properties such as interfacial energy and mobility, is a major issue in phase-field simulations. The ultimate goal of this series of studies is to develop a method that integrates the PF method and in-situ observation using data assimilation. Here, as a preliminary evaluation of the data assimilation using in-situ observation data, we investigate a data assimilation system for obtaining material properties using twin experiments of directional solidification of a binary alloy.

Machines such as jet engines and gas turbines are required to reduce the emission of carbon dioxide. In order to achieve this goal, the temperature at the turbine inlet and other parts of the turbine have been raised to a high temperature in order to increase the efficiency of these machines. However, creep damage has emerged as a problem as a result of higher temperatures. Creep damage is known to reduce the grain boundary strength by decreasing the crystallinity near the grain boundary, resulting in brittle grain boundary fracture. We have confirmed the local energy concentration near the grain boundary and the atomic exhaustion behavior at the grain boundary by the creep loading analysis using the molecular dynamics method with the bi-crystal model. In this study, the effects of disordered atomic arrangement and lattice mismatch near the grain boundaries on the crystallinity degradation phenomenon were clarified by creep loading analysis using a bi-crystal model with a combination of different crystal orientations.

In this study, topology optimization based on effective energy concept is applied to the optimal design of compliant mechanisms. In the proposed formulation, the effective energy, which is the ratio of input energy to output energy, is maximized as the objective function. This formulation eliminates the need for a virtual spring element at the input port because the stiffness to the input force is ensured by minimizing the input energy of the objective function. The proposed method enables us to design a compliant mechanism in which no reaction force is returned from the output port. Furthermore, by changing the values of the parameters α and β in the objective function, it is possible to control whether the optimization prioritizes minimizing the input energy or maximizing the output energy. In other words, by adjusting these parameters, it is possible to control the properties of the structure obtained by optimization. Several numerical examples are presented to confirm the effectiveness of the proposed method.

The backwash of a large number of smart houses equipped with photovoltaic power generation systems has led to a decline in the equipment utilization rate of power plants that are interconnected with the commercial grid. A Home Energy Management System (HEMS) that uses both photovoltaic power generation and storage batteries is effective in improving these problems, but highly accurate solar radiation prediction is necessary for optimal operation. In this research, we will create a model for predicting solar radiation using LSTM (long short-term memory), which uses environmental information and sky images as feature value, and investigate the significant differences in each feature value.

In automobile collision analysis using the finite element method, the uniformity and orthogonality of mesh pattern influence analysis accuracy, which causes variation of the pattern, correction cost and reduction of analysis accuracy. Multiple workers often make visual judgments manually to maintain the quality of the mesh pattern. In this study, we propose a method to evaluate the quality of mesh pattern using GCN (Graph Convolutional Network), which is a graph-based machine learning method. As training data, we created adjacency matrices for the finite elements form and converted them into graphs. These meshes were previously labeled by an experienced operator. We trained mesh patterns and made inferences for the unknown data using the labeled meshes. The accuracy could be improved from 49% to 63% by giving the nodes coordinate features, and the accuracy of 3D meshes was only 23%.

This paper proposes an equivalent circuit model which includes the effect of displacement and eddy currents. In the proposed circuit model, the AC resistance and complex inductance are introduced to represent the skin and proximity effects. In addition, the inductance and capacitance are evaluated from the static electromagnetic field computation. To verify the proposed circuit model, the simple inductors with and without magnetic core are analyzed by the proposed and conventional circuit models, and those numerical results are compared.

If available learning data are not enough, there is an issue that the prediction accuracy by machine learning become deteriorated. For example, in the case of tire properties prediction, even if the prediction accuracy is not sufficient by learning with past real data, specifications and measurement data, it was confirmed that the prediction accuracy can be improved by transfer learning with source machine learning model using virtual data generated by finite element method.

In automobile manufacturing, friction stir welding (FSW) is often used to join aluminum materials, but they are deformed depending on the joining conditions. Therefore, prediction of the deformation of them using simulation is required. FSW is a complex physical phenomenon such as tool rotation, tool movement, material softening due to heat generation, and plastic flow, which is difficult to simulate in detail. In this study, we investigated a simple and high-speed method for predicting the deformation caused by FSW using the inherent strain method. In addition, by adopting the Hierarchical Domain

Decomposition Method (HDDM), large-scale analysis was speeded up to put into practical use. As a result, it was found that the inherent strain can sufficiently reproduce the deformation and the effect of longitudinal contraction is remarkable. It was also found that it is possible to use this method in practice by improving the computation the speed.

Nickel-chrome plating is a typical anti-corrosive under atmospheric surroundings. But, when a pinhole defect reaches the region of nickel or iron, pitting corrosion will occur along the defect. In order to simulate pit growth across the interface between regions of nickel and iron, it is necessary to consider different polarization characteristics of nickel and iron, which is tricky due to the moving of the interface during pit growth. In this work, we proposed a method to realize automatic switching of different polarization characteristics when the corrosion penetrates through the interface between the regions of nickel and iron using the Phase-field method. The pitting corrosion from nickel to raw iron is analyzed and the computational efficiency of Phase-field method is examined.

The phenomena of emerging stresses on a moving plate were calculated using the general-purpose mechanical analysis software Adams, MSC. The simulation results were verified by an experiment in which strain gauges were attached to the plate which slides along with a rail. The simulated stresses emerged at the plate were close to values measured by the strain gauges in the experiment. Also, geometry optimization was conducted using the general-purpose optimization software OPTISHAPE-TS,Quint. The geometry optimization not only improved stiffness of the plate, but also decreased the mass of the plate in simulation and experiment.

In order to improve the mechanical reliability for wind turbine drivetrain structure subjected to random dynamic load, market load and fatigue calculation method is developed based on stochastic and stochastic methods (extreme statistics) of the mechanical composition part to the random dynamic load due to the actions such as wind and the earthquake. Evolutionary spectrum is introduced from the analogy with a concept of a dynamic design wave. Random dynamic load calculation and fatigue prediction of 2MW wind power generation plant by the strong wind and earthquake are attempted using large-scale structural analysis based on Finite Element Method and Surrogate Modeling Method based on machine learning with physical model.

In this study, we investigate the effect of surface energy on the nano-scale object using the theory of second strain gradient elasticity. This theory adds the first- and second-order gradients of strain to constitutive equations of the classical elasticity. The characteristic length scales, which represents the size of the microscopic structure of the material, and the surface energy are introduced into the constitutive equations. We solve the variational problem by using the isogeometric analysis; Galerkin method with non-uniform B-spline basis function. Strain energy density is simplified by using the irreducible decomposition of elasticity tensors under the general linear group GL(3). We conducted a numerical analysis for a hollow torus-shaped isotropic elastic medium. We observed that the surface displacement tends to dominate as the size of the material approaches the nanoscale.

Cohesive Zone Model (CZM) is often used to design adhesively bonded structures by CAE. To establish the measurement method for identifying traction-separation law of CZM directly with a versatile apparatus, SAICAS (Surface And Interfacial Cutting Analysis System) with elaborated blade is studied. Shear test is performed by horizontally pushing adhesive brick on adherent while measuring reaction force against the blade and displacement of the blade. Traction-separation shape shows rectangle in the case of cohesive failure and triangle in the case of adhesive failure. Regarding triangle shape of traction-separation law, SAICAS simulation is performed by setting identified traction-separation law by SAICAS experiment to cohesive element on interface of adhesion and adherent. Traction-separation law obtained from SAICAS simulation correspond well with that of SAICAS experiment. It means proposed method can identify triangle shape of traction-separation law for mode II interfacial fracture of bonded joints. This method is also useful for screening test of a small quantity of adhesive on early stage of development, application examination of adhesives on special adherent, evaluation of adhesive property by specimen from cut-out component after durability test and so on.

Offshore wind power generation are placed offshore as fixed/floating structures in ocean. Therefore, it is necessary to ensure the stability of the structure and the safety against waves. It is important to establish an appropriate simulation technique. Various types such as spar type, semi-sub type and others are used for floating offshore wind power generation. We used the particle method to simulate the behavior of a spar type floating offshore wind power generation which has high stability. The δ-SPH method which is proposed based on the SPH method is used in the analysis. The δ-SPH method is a method in which an artificial viscosity term is added to the equation of motion and an artificial density diffusion term is added to the continuity equation in order to suppress the numerical oscillation that appears in the general SPH method. In this study, numerical analysis is performed using general SPH method and δ-SPH method respectively, and the results are compared to confirm the usefulness of the δ-SPH method.

In railway ballasted track, the settlement of a ballast layer is caused by accumulation of irreversible displacement induced by train passing. The ballast settlement can be modeled using the elastoplastic FEM with the the cyclic densification model. The railway ballast, which is an assembly of crushed stone, may have spatial variation of elastoplastic behavior. The influence of spatial variation of several elastoplastic parameters on the simulation results is evaluated using the Stochastic FEM (SFEM) with the stochastic collocation method.

We have been developing a parallel partitioned simulation system to solve large scale real world’s coupled problems. This paper describes some key technologies of the system, and introduces its application to multiphysics combustion simulation of lab-scale coal gasification reactor.

The multi–phase–field (MPF) method is a promising interface tracking method for multiphase flow problems with three or more phases. We have developed the MPF model using conserved Allen–Cahn (MPF–CAC) equation [Comput. Fluids, 178, 2019, 141-151]. However, this model causes a slight volume change around multiple junctions and lacks extensibility to more complicated multiphase flows. In this study, we develop a new MPF–CAC model with a better conservation property by improving the previous MPF–CAC model. The volume conservation property of the new MPF–CAC model is evaluated by comparing to the previous MPF–CAC model and Cahn–Hilliard model.

Nonlinear finite element analysis with deformed surface constraint generated by virtual point cloud was performed. We created a virtual point cloud of a deformed beam, and triangle polygons as deformed surface. The beam created for constructing deformed surface constraint was used as a reference solution. We analyzed another beam that had same mesh but different distributed load with deformed surface constraint to match the deformation to the reference solution. The calculation was converged in about 4 minutes and the accuracy of displacement was mostly in good agreement. However, Cauchy stress included large error especially near the free end of the beam.

In this research, we proposed multi-layered core panels based on a truncated octahedron in order to develop a core panel with quasi-isotropic impact absorption performance. The core panel was designed by tessellating a truncated octahedron in plane and then layered to the out-of-plane direction. The results of dynamic analysis revealed that increasing the number of layers in the multi-layered core panel can increase the proportion of plateau region of reaction force during the in-plane and out-of-plane deformations so that the specific energy absorption was improved. In addition, since the truncated octahedron has symmetry in shape, it was confirmed that the amount of absorbed energy in the out-of-plane direction was approximately equivalent to that in the in-plane direction. This finding shows an advantage of the proposed core panels because they can solve the drawbacks of conventional core panels that the stiffness of the panels is anisotropic and the impact absorption performance is highly affected by impact directions.

In this study, a three-stage folding blast demolition planning of buildings was introduced by using Key Element Index (KI) which can numerically evaluate the contribution of a column to the strength of buildings. In addition, to improve the efficiency of blast demolition, time interval between the second and the third blast was selected according to the impact load of the upper part of the building acting on the lower part after the second blast. The Adaptively Shifted Integration (ASI) - Gauss code which can stably perform non-linear structural analysis was used as a numerical code to simulate and investigate the blast demolition plan. From the numerical results, the relationship between the number of removed columns and the height of remains, the scattered distance after blast demolition were investigated to see the effect of time interval between the blasts on the efficiency and safety of the folding blast demolition of buildings.

In order to promote the effect of a thrombolytic agent as a non-invasive cure for the cerebral infarction of an acute stage, a real-time adaptive ultrasonic lens constructed by arranging point-like scatterers on a lattice has been proposed. The structure was optimized from the initial arrangement of the scatterers determined by binarizing the interference pattern of the radiative wave from a point sauce at the focal point and the reference plane wave with the analogy of holography. The dependence of convergence of optimization on the initial state and on the thickness of lattice has, however, never been investigated, yet. In the present paper, four kinds of the initial states are considered, that is, the holographically generated configuration, the configuration without the scatterers on all of lattice points, the configuration with the scatterers on all of lattice points and the configuration with the randomly distributed scatterers. The convergence of optimization for three kinds of thickness, that is, the number of layers is also systematically investigated. It has been found that the initial configuration without the scatterers shows the fastest convergence and increase of the number of layers improves the ratio of the power of the ultrasound wave at a focal point to that of the background.

We perform the prediction of flow fields with two steps in a duct from learning of ones with only one step using convolutional neural networks (CNNs) and signed distance function (SDF). We show that CNN with U-Net architecture is able to predict the flow fields with two steps. Furthermore, we find the main factor which enables the prediction of the flow fields with two steps is to use a CNN architecture without fully connected layer, from the comparison with the unpredictable network.

For the production of future and finer machine elements, it is important to understand how plastic deformation behavior of Mg polycrystal is affected by its microstructural changes, since this material has strong anisotropy on the deformation. In this study, we construct several molecular dynamics models of Mg nano-polycrystals including different types of grain boundary (GB) and conduct uniaxial tensile testing to analyze their microstructural plastic deformation behavior, such as twinning, dislocation slip, and structural change in GB region. As a result, it is confirmed that anisotropy of Mg polycrystal appears before the yielding of the whole specimen, even for those nano-polycrystalline models that have high density of GBs. In addition, propagation and multiplication of dislocations are observed during plastic deformation around an initial edge dislocation occurred in GB region and around the triple junction of GBs.

Avoiding vibration during processing is an important issue for industry. A numerical approach to the prediction of chattering vibration requires several numerical models, including dynamic models of the workpiece and tool, cutting interaction laws, and surface representations. The resulting model is a trade-off between the complexity of the previously mentioned elements in the solution strategy for a given equation of motion. This paper describes a case study of predicting chattering vibration in milling using an instantaneous contact force model, which is one of the methods for cutting force simulation.

With the recent intensification of disasters, as exemplified by the 2011 off the Pacific coast of Japan earthquake, the need to strengthen existing damage countermeasures and damage estimation has become increasingly important. Although simulation is an effective method for damage estimation, it consumes enormous computational resources because the relevant parameters need to be determined empirically or by human iteration. In this study, we employed PINNs (Physics-Informed Neural-Networks), one of deep learning framework, to perform inverse analysis of parameters from observed physical quantities. PINNs is a deep learning framework in which physical validity is ensured by adding physical constraints such as governing equations, initial conditions, and boundary conditions. These properties are expressed by the two items "prediction loss" and "physical loss", which are unique to PINNs. "Prediction loss" evaluates the error between the observed data of each physical quantity and the physical quantity predicted by PINNs, while "physical loss" evaluates whether the predicted value of PINNs satisfies the governing equation. In this study, we examined how reliable the PINNs are based on observation data containing strong noise, which is expected in practice. As a result, it was shown that the accuracy of parameter estimation was significantly reduced when strong noise was included in the observed data.

We propose a computational experiment method for granular materials applying the renormalization molecular dynamics (RMD) method. Previous reports have shown that shape models using the signed distance function (SDF) can accurately predict the amount of materials carried by a screw feeder. However, in the analysis of a thermoplastic screw, it is difficult to analyze the whole screw because resin pellets melt and flow after the feed section. In addition, when performing a partial analysis of only the feed section, it is necessary to consider the subsequent back pressure. In this paper, we propose boundary conditions to consider the back pressure. It is shown that the extrusion amount can be predicted by analyzing the plasticized screw feed section using this boundary condition.

In computational analysis for living tissue such as blood vessel, meshes are generated by medical image and it depends on quality of medical image. The purpose of this research is to verify directional dependency on medical image data by k-refinement.

A methodology to evaluate the J-integral for Isogeometric analysis (IGA) is presented in this paper. A domain integral method that is suited for IGA is proposed. The results of example problems show favorable comparisons with reference solutions.

CAD is required to execute CAE, but not all models of CAD exist at the design site. Even if CAD exists, it will be difficult to generate a mesh unless the fine geometry that is not necessary for analysis is removed and the model is made as simple as possible. In addition, considering the application to shape optimization, it is necessary to automatically create CAD that reflects the arbitrarily changed shape, which further raises the hurdle. On the other hand, we noticed that design tools such as Excel, which are used at the design site, sweep out point clouds data for shape recognition. If a triangle is created on the surface by connecting these point clouds appropriately, it will be CAD in STL format. In this paper, we report on the development of a 3D modeler that automatically creates STL format CAD from points cloud data.

One of the attempts to achieve high ductility of materials is the sequential activation of multiple deformation modes. The sequential activation of deformation modes, such as dislocation motion, deformation twinning, and phase transformation, improves work hardening ability and increases ductility. In this study, a binary atomic model with various deformation modes is constructed by adjusting the parameters of the interatomic potential, and the possibility of sequential activation of deformation modes under tensile deformation is investigated by molecular dynamics.

The Lattice Boltzmann Method (LBM) is a method suitable for parallelization and has been attracting attention with the recent increase in computer speed. In this study, the LBM has been applied to the topology optimization problem in the field of thermo-fluids. In the previous topology optimization of thermo-fluid field using LBM, the thermal conductivity of fluid and solid was assumed to be the same. In this study, the problem of minimizing the average temperature of a heated surface under natural convection is treated as a concrete example, and the sensitivity analysis and geometry optimization based on the accompanying variable method was conducted by considering the difference of the thermal conductivity for solid and fluid parts.