With the increasing concern for environmental issues, biodegradable plastics are expected to be widely used. However, disadvantages like poor heat resistance and relatively low strength limit the application scope. This study focuses on the crystalline phase of poly (L-lactic acid) (PLLA), a bioplastic that has attracted much attention. Molecular dynamics (MD) simulation is used to investigate the effect of microstructure changes on mechanical properties, thereby seeking methods to improve the mechanical properties of PLLA. In this study, the model of the PLLA crystalline phase is established, and uniaxial tension and compression simulations are performed after relaxation. The following conclusions are drawn by analyzing stress-strain curves and structure changes of molecular chains during deformation: The mechanical response of the crystalline phase of PLLA is anisotropic. In compression, strain softening occurs after the buckling of molecular chains when the strain direction is parallel to the chain axis. For the vertical strain, the slipping of molecular chains is one of the reasons for the strain softening behavior during deformation.

In recent years, composite polymers have been studied to improve the mechanical properties of polymeric materials by adding hard particles to adjust their properties. In the development of composite polymer materials, the elucidation of mechanical properties at the molecular scale is a challenge. In this study, a simple model system for both polymers and hard particles is used to understand the effect of hard particle addition to crystalline polymers universally. A composite polymer model is constructed by adding CNTs to a crystalline polymer, and coarse-grained molecular dynamics simulations are performed to analyze the friction and wear behavior on the polymer surface. The friction phenomena in CNT-free and CNT-containing polymers were compared by sliding metal spheres on the surface. The Lennard-Jones potential was used for the interaction between the polymer and the metallic sphere. Simulations showed that without CNTs, the metallic spheres penetrated the polymer during the friction process and eventually reached the bottom of the model; for the CNT-doped polymer, no penetration of metallic spheres into the polymer was observed during the friction process. CNT migration is little when the metallic spheres and CNTs come into contact during friction. Therefore, the presence of CNTs at the friction interface prevents the wear of the polymer during the friction.

Coarse-grained molecular dynamics (CGMD) simulation has been frequently used to analyze the physical properties of cross-linked polymers. The effects of cross-linker functionality and distribution on the stress−strain behavior of cross-linked polymer networks are studied using CGMD simulations. Stress−strain curves are determined for each system from tensile stretching simulations. The radial distribution function and Voronoi analysis are also used to study the structures of the simulated systems. It is notified that the bridge point distribution plays an important role in the change of the stress-strain curve.

In this paper, influences of how the local patch is set in the s-version Isogeometric Analysis (s-IGA) are discussed. In s-IGA, the problems domain is modeled by the global IGA patch. Local features such as cracks, holes, etc. are represented by the local IGA patches. The local IGA patches are superimposed on the global IGA patch. The assumed displacements are set by the sum of those of the global and local patches in the domain of the local patch. It is more tractable to generate the local features by the local patch only than by the monolithic global patch. To assure the accuracy, the size and descretizations of the local patch relative to the elements of the global patch are considered to be important factors. Some considerations on their influences are reported for two dimensional linear fracture mechanics analyses.

When the finite element method is used in structural analysis, the problem of extremely low analysis accuracy due to poor quality mesh partitioning has been a concern, and mesh-free analysis methods that do not depend on mesh partitioning for calculation accuracy have been proposed. The fragile Points method (FPM) has been proposed as one of the mesh-free analysis methods suitable for crack propagation analysis, and FPM has the advantage that crack propagation analysis can be easily performed without remeshing the internal boundaries of the model. In this study, the accuracy of FPM in fracture mechanics analysis was verified using an infinite flat plate with Mode I cracks.

An implicit algorithm in which a primal-dual interior point method (PDIP method) is applied for Gurson-Tvergaard-Needleman model (GTN model) is presented to stabilize the stress update. Although the GTN model widely utilized to realize the change of void volume fraction that dominates ductile fracture in metals, the numerical instability occurs due to the shrinkage of yield surface and the acceleration of void growth. In particular, the conventional return mapping algorithm leads to the misjudgment of yield condition since the yield surface shrinks by the evolution of void volume fraction. In addition, the smoothness of solved equations is required to employ the nonlinear solution method such as the Newton method, whereas the evolution of void volume fraction is approximated as bilinear form to represent the acceleration of void growth. Against these backgrounds, we apply the PDIP method for the stress update of GTN model, in which the inequality constraints in the constitutive model are replaced by an equivalent constrained optimization problem to ensure the numerical stability. Finally, the capability of our proposed PDIP method is demonstrated throughout several numerical examples that cannot be solved by the conventional return mapping algorithm or the PDIP method applied for only yield function.

The contribution of this study is to propose a cohesive traction embedded constitutive law combined with shear-induced damage and kinematic hardening law to represent the failure of metallic materials subjected to cyclic loading under various stress states. The proposed law accommodates a hyperelasticity-based plastic model with the use of the deformation gradient multiplicatively decomposed into separation-induced, elastic parts and plastic one that is further decomposed into energetic and dissipative ones to realize nonlinear kinematic hardening under cyclic loading. Moreover, to realize both flat fracture and shear-lip fracture under various stress states, both cohesive traction separation law and shear-induced damage are incorporated into the hyperelasticity-based plastic model. The stress release process along with the material separation due to void nucleation, growth and coalescence under high stress state is represented by the combination between the separation-induced deformation gradient and the cohesive traction separation law. On the other hand, the shrinkage of yield surface that is caused by the rotation and elongation of voids in a shear band under low stress state is realized by the introduction of the shear-induced damage into the Tresca yield function. The capability of our proposed constitutive law is demonstrated by comparison with experimental results under various stress states including monotonic and cyclic loading.

The life evaluation method under very low cycle fatigue is still under research. Existing evaluation parameters cannot be applied because of constraint effects and large plastic deformation. In a previous study, a crack growth condition was proposed and evaluated using the equivalent plastic strain amplitude and stress triaxiality at the center of the plate thickness. Such a physical quantity cannot be directly obtained by an experiment. In this study, numerical simulation of crack propagation with three loading conditions were conducted. The evaluation method was based on substituting the physical quantities of the individual crack front edge positions at each load condition into the crack propagation condition equation, which was not applicable to the surface area. Therefore, the evaluation was modified to take into account the difference of stress states between the center of the plate thickness and the surface area. Crack propagation was not observed in the intermediate area between the center of the plate thickness and the surface area.

In this study, a stress analysis method based on the extended finite element method (XFEM) to effectively model cracks in thin-walled lightweight structures is developed. More specifically, a method based on XFEM using continuum shell elements, which can model cracks independently of elements, is proposed. In the method, the crack is expressed implicitly by the level set method using two kinds of signed distance functions, and a cohesive zone model is introduced to the crack to model matrix damage. This paper shows a formulation of enriched continuum shell elements, an outline of the developed code referred to as NLXSC8, and numerical results for verifications.

In this study, elastic-plastic FEM using cohesive zone model with trapezoidal traction-separation law is applied to ductile crack propagation analysis of a C(T) test specimen and the relationship between load-line displacement (LLD) and applied load is evaluated. Crack extension Δa is determined by damage variables of interface elements, and ΔaLLD relationship can be obtained. In addition, it is possible to assess the J-R curve by conducting several of stationary crack analyses using the Δa-LLD relationship.

In recent years, composite materials such as Fiber Reinforced Plastics (FRP), which are lightweight and have excellent strength, have been attracted attention. For the application of composite materials to structures, it is essential to understand the orthotropic fracture properties of the materials. However, fracture phenomena in orthotropic materials have not been analyzed in many cases compare to isotropic materials. In this study, fracture mechanics analysis using adaptive strategy and wavelet Galerkin method for composite materials is proposed in this research. In numerical example, fracture mechanics for 2D plate is analyzed to verify the proposed approach.

Automatic flow visualization methods are being pursued. Flow visualization has a long history and many useful visualization methods have been developed. However, because of the growing scale of computer simulations, flow visualization tends to be a routine process and various flow visualizations are difficult to apply with the limitation of computer cost. We have been developing automatic visualization method including all flow features in one shot of the image using Self-Organizing Map (SOM). SOM has a function to reduce the dimensions and various flow functions are possibly degenerated down to two or three dimensions for visualization. In this paper, the present method is applied to the three dimensional flow and its validity and usefulness are discussed.

This research aims to use AI and machine learning to classify music genres and assist artists. By objectively analyzing their own music, artists can better understand the characteristics of their music and incorporate popular trends. They can also evaluate their own work, set goals, and drive continuous improvement. In this work, we use Music2vec as a method to convert music into vector representations for classification. The study plans to classify music by genre, country of origin of the artist, and songs that are popular in each country. Experiments using Music2vec and the GTZAN dataset achieved a classification accuracy of 62%, demonstrating the feasibility of classifying music genres. However, further research is needed to explore the subdivided genre taxonomy and the impact of the artist's country of origin on the taxonomy.

We aim at developing a Physics-Informed Neural Network (PINN) to estimate eigenfrequencies of 3D elastic bodies rapidly and accurately. As the first step, we consider a Convolutional Neural Network (CNN), which is based on the VGGnet. The CNN is trained by a dataset of geometries, volumes, and (first-order) eigenfrequencies of elastic bodies. The relative error of the estimated eigenfrequency was about 15% and 30% for the training and validation data, respectively. This result indicates to consider a PINN by introducing eigenequations to the loss function, for example.

We apply the framework of shape-optimization, which minimizes/maximizes a given objective function with respect to latent variables in terms of variational autoencoder (VAE), to three-dimensional unsteady acoustic problems. Since a VAE is trained so that it can output similar data to the input data, the output structure for every latent variable can succeed one or more characters (outlooks) of the input structures. We confirm that out VAE can produce similar 3D objects to the input ones. Then, we perform some shape optimizations to create an optimal structure that can focus sound field to a certain point by using the latent variables as the design ones.

As an evaluation method for concrete structures, information from hammering inspections can be digitized by acoustic emission sensor. The state of structure is predicted by empirical condition or machine learning using the information. The measurement system is called as digital hammering inspection. In this study, XGBoost, which can present the importance of input factors, was used to predict the state of concrete structures. The measurement results of hammering inspections of concrete specimens were used as training datasets. The predictions were made for four types of concrete structures: reinforced concrete, spiral sheathing, foamed concrete and unreinforced concrete. Accurate predictions were made by changing our data acquisition method based on the importance of the input factors and the clustering analysis.

Digital hammering inspection using acoustic emission (AE) sensor is expected to be a method to inspect concrete structure. Digital hammering inspection using AE sensor measures vibration waveforms using mechanical resonance of piezoelectric elements. In this study, the embedded hardware, which is a structure consisting of a rectangular plate with multiple headed round bars (headed shear connector) attached to it, is evaluated using the vibration waveform. Due to difficulties of noise treatment and instability of the measurement, machine learning techniques, which are random forest and XGBoost, learn and predict. Input parameters were made from the waveform and the spectrum by FFT. Features which can improve prediction accuracy were selected based on the importance of decision tree. Binary classification is used to predict whether the headed shear connector is cut or not. In this study, methods for evaluating the embedded hardware and feature selection methods to improve prediction accuracy are discussed.

The development of automatic tooth region extraction from cone-beam computed tomography (CBCT) using machine learning is expected to promote dental treatment digitalization. Machine learning requires many training datasets, however, it is difficult to prepare many training data sets in medical data, in addition, the preparation of training datasets takes a huge effort. Data augmentation is a method to prepare many training data sets from a small number of training datasets. This research examined the effectiveness of data augmentation methods on the accuracy of tooth region extraction. The examination suggested that the rotation approach improves the accuracy of tooth region extraction.

Element stiffness matrices in the FEM are usually calculated using Gauss-Legendre quadrature, and it is well known that the accuracy of the stiffness matrix of an element depends on the shape of the element. Deep learning can help to improve the accuracy of the stiffness matrix, and a new method to predict optimal quadrature parameters to be used for each element by deep learning has been proposed. Using optimal quadrature parameters for the numerical quadrature of an element, one can expect much better accuracy for the stiffness matrix. In this paper, optimal quadrature coordinates are obtained using the steepest decent method, and their characteristics are investigated in detail.

In recent years, CAE analysis reduces a number of experiments and costs in the design process of manufacturing. However, the computational cost of CAE analysis has not decreased due to the increased number of design trials. Although machine learning predictions are less accurate than CAE, they are expected to be 100 to 1,000 times faster, and can investigate almost any desired solution in a very short time. In general, it is difficult to predict extrapolation in regression problems. Rules based on physical phenomena should be well-trained in the predictor for extrapolation. Therefore the predictor is required to improve generalization performance and adapt to engineering problems. The learning method uses sub-voxels, which are containers of local physical quantities and material properties, as input parameters. We utilize 2 dimensional and 3 dimensional convolutional neural network (CNN) for regression prediction. In this study, we proposed a learning method in which the input parameters are sub-voxels, and evaluated the prediction accuracy of a predictor composed of 2D-CNN and 3D-CNN．

In this study, the accuracy of plural crack propagation prediction was improved by machine learning with consideration of physical quantities using a small data set. A dataset is obtained from the results of crack propagation analyses using s-version FEM combined with an automatic mesh generation technique. The input parameters are the coordinates of the four crack tips. The output values to be predicted are crack propagation vectors and a number of crack propagation cycles of 0.25mm length. Crack propagation paths and rates are predicted with less 1/300 error than the previous study. The output parameters, number of cycles, and crack growth vector, are different in the way variation and the magnitude of the values. Individual networks for each value to be predicted can keep a uniform distribution of activations in each layer. In this way, problems such as gradient loss and limited expressiveness can be avoided. As a result, it is possible to predict the shape and a rate of crack propagation within 0.07 percent relative error.

Evaluation of displacement and strain of structures and members is adequate for evaluating their soundness. Full-field measurement using digital images for this evaluation became widespread. On the other hand, the accuracy of the full-field measurement depends on the measurement environment and spatial resolution. Displacement field prediction using deep learning techniques can be a practical approach for alleviating this problem. To achieve this objective, we have developed a displacement field prediction method based on the theory of deep energy method. As a result of a previous study, the learning rate dramatically affects the improvement of prediction accuracy was confirmed. In this study, we investigated the effect of the learning rate on predicting the displacement field using the developed method. The initial and final values of the learning rate and the number of learning times, which are learning parameters, were adjusted. We discussed the prediction accuracy of the displacement field using these parameters.

Generally, the learning data is limited in predicting stress or strain field numerical analysis results using deep learning techniques. Data augmentation techniques such as noise addition are effective in improving prediction accuracy. However, in noise addition is necessary to consider the added noise carefully. The proposed data augmentation method that equalizes relative frequency augments the data efficiently, and it is possible to improve prediction accuracy with fewer learning times. In this paper, we applied the proposed data augmentation method to generate the learning model for displacement field prediction. This learning model used two-dimensional finite element analysis results in different loads with the same boundary condition. Additionally, we predict displacement fields using this model and learning of stress and strain fields using an automatic differentiation technique in the deep learning library. We predict the stress and strain fields of an analysis result not used to generate two learning models and discuss the results.

A determination of voids and cracks in materials simultaneously using topological derivatives is considered. We determined voids using topological derivatives for voids and determined cracks using topological derivatives for cracks in our previous studies. In this study, we determined both voids and cracks in materials simultaneously using topological derivatives for voids and their gradients in a simple problem.

This work presents a fast iterative solver for 2D Helmholtz' transmission problems, based on skeletonization, a technique commonly used in fast direct solvers. The skeletonization-based iterative technique utilizes the explicit form of hierarchical off-diagonal low-rank matrices to efficiently perform adjoint matrix-vector multiplication. This approach enables the use of a variant of IDR(s)Stab(l), which was difficult to implement with the fast multipole method. Numerical examples demonstrate the superiority of the proposed method, which combines IDR(s)Stab(l) and skeletonization, over the combination of ordinary Krylov subspace methods such as BiCGStab or GMRES with skeletonization.

This study proposes an inverse analysis method to estimate the position and shape of scatterers from profiles of sound pressure measured at prescribed observation points. We describe each scatterer with NURBS patches and manipulate their control points (CPs). Then, we cannot move each CP largely because the resulting CPs can cause self-intersections of the NURBS patches. To avoid the issue, we estimate the position and shape of each scatterer sequentially. Further, we perform a global search of the initial position of each scatterer at first. This can avoid the to fall in a local optimal solution.

This study proposes a time-domain boundary element method (TDBEM) for the 3D dissipative wave equation. The numerical experiments reveal the proposed TDBEM, which is based on the ordinary boundary integral equation and the time-marching scheme, works accurately for Neumann problems.

In Japan, as the population ages, the number of cases of Alzheimer's disease is increasing. Alzheimer's disease has traditionally been regarded as a disease of the nervous system, but it has been revealed that it is caused by arteriosclerosis due to the deposition of amyloid β in the capillaries of the brain. It has been found that amyloid β is discharged by stimulating cerebral blood vessels with weak ultrasonic waves. Treatment with that method is about to enter Phase 3 clinical trials. Existing devices for other applications are, however, used for brain stimulation, and the sound field used for stimulation has never been optimized. In this study, we propose to uniformly stimulate the brain with a single irradiating plane wave using a transcranial lens. We have already proposed transcranial lenses with appropriately placed point-like scatterers. By means of modifying the procedure to optimize the arrangement of point-like scatterers, the optimal arrangement has been determined so as to focus the incident plane wave to have uniform intensity within the brain.

Finite element method (FEM) with periodic conditions is widely used for designing topological phononic crystals (PnCs), which enable to fabricate robust waveguides against disorder and defects. However, it takes a lot of time to calculate dispersion relations with FEM analysis because calculation at a lot of wavevector points is necessary. In this study, we use tight-binding (TB) approximation for modeling 3-dimensional PnCs which have topological hinge modes. It is revealed that TB approximation can estimate dispersion relation of 3-dimensional PnCs, forming stacked Kagome lattice, with FEM simulation precisely around K point and H point. Moreover, we show that numbers of degrees of freedom decrease from tens of thousands to three with this approach, which reduce computational time. The parameters of TB model we propose are corresponding with the parameters of FEM simulations, thus this model enable to design topological elastic devices instead of FEM simulation. Finally, we discuss with estimating dispersion relation of supercell and topological invariants.

Phononic crystals have attracted much attention in recent years for controlling the acoustic properties of materials. This property is achieved by controlling acoustic waves in a specific frequency band called a band gap. However, identifying phononic crystals with a band gap in a specific frequency range is a computationally expensive process. By applying inverse problem approach using multilayer neural networks, we propose an approach to predict the materials and structures of one-dimensional phononic crystals with the desired bandgap, offering an efficient method for the design of phononic crystals.

Graphene is a nanomaterial with excellent properties such as high electron mobility, and is expected to be used in high-speed electronic devices. While the degree of the controllability for electronic structures is indispensable in device design, but graphene is a difficult material to implement such a feature as it has an atomically thin sheet-like structure. In this study, we propose an electronic-structure control using a Moire structure in which two layers of graphene are rotated and laminated with each other. In the Moire structure, it is possible to vary the interlayer interaction by changing the rotation angle and the interlayer distance. We aim to design an electronic device where electron distribution on defects can be actively controlled via the interaction of defects in a graphene with Moire structure. Using a tight-binding model we demonstrate that the electron distribution around the defect in the ground state can be controlled by the Moire structure. The degree of localization around the defects for the electrons injected in the lowest conduction band can also be varied by the interlayer interactions.

In order to evaluate the risk of fatigue crack initiation from welded joints, it is necessary to consider the effects of residual stress and weld defects such as cracks. Finite element analysis is an effective tool for evaluating the stress intensity factors of a crack under the combined stress field with arbitrary shaped structure. In this study, in order to apply the thermal-elastic-plastic analysis to a large-scale real structure, we tried a simulation that reduced the computational cost by limiting the modeling range. As the result, the analysis time was reduced to149 hours (6.2 days), and the difference from the measured value was about 50%. It was confirmed that the stress intensity factor in the combined stress field can be calculated by applying the residual stress values and static analysis results to the crack surface loading method based on the principle of superposition.

Unexpected leakage can occur at the fastening points of various industrial products. The proactive evaluation at the development stage is necessary to prevent leakage of gases and liquids during product use. Sealing can be evaluated by calculating contact surface pressure using the finite element method (FEM) and by calculating the amount of leakage with contact theory, considering micro geometry. In general, these methods do not consider effects of machining marks, although which influences to amount of leakage. Therefore, this study proposes a sealing evaluation method to consider the effects of machining marks by combining the FEM and Persson's contact theory. The influence of the contact surface shape is verified by elemental tests, and it is confirmed that not only the roughness but also the machining shape had a significant influence. The same roughness, however, the surface pressure direction intersects the machining shape, the leakage can be 10 times larger than non-intersects one. It is also found that the amount of leakage increases exponentially with roughness. This sealing evaluation method effectiveness confirms that the actual measurement of leakage.

Ti-6Al-4V is widely used for aircraft engine parts due to their high specific strength. It is known that micro texture region, in which crystal orientations are locally concentrated in a specific direction, in Ti-6Al-4V causes a decrease in life due to cold dwell fatigue (CDF), where fatigue and room-temperature creep are superimposed. Therefore, we investigated the fracture mechanism and the life prediction method for CDF by calculating the stress and strain in and around the micro texture region by the crystal plasticity finite element method (CPFEM) and digital image correlation (DIC).

Wiredrawing is a conventional plastic working method to produce metal wires, but a further improvement in strength and functionality of wires can be achieved by producing finer wires with nano-meter diameter. In particular, ultra-fine wire made of magnesium (Mg) and its alloy is expected for industrial purpose, including medical equipment. However, grain boundaries and crystal defects such as dislocations and twins in Mg and Mg alloys will significantly influence overall performance of such ultrafine wires. We are clarifying the deformation mechanism, considering the effect of grain size. Conventional wire drawing conditions were applied to a nano-sized pure Mg model. Molecular dynamics (MD) simulations with EAM potentials and simulations were performed by changing the grain size. In the results, the grain rotation due to twinning occurs so that the <1010> direction is coincided with the drawing direction. This result agrees well with the experimental texture obtained in the wiredrawing of Mg. In conclusion, wiredrawing process in nano-polycrystal includes little basal slip deformation, but instead twinning and grain boundary sliding mainly occur. We observed that a certain texture is also developing in the nano-polycrystals.

In the boundary lubrication, where both fluid lubrication and solid contact states coexist, the formation and delamination of lubricant layers on the solid surface are expected to occur alternately. This study aims to reveal the nanometer scale mechanisms of lubricant layer formation and delamination by means of coarse-grained molecular dynamics simulation. Our calculations demonstrated time-dependent transition from the slip between lubricant layers to the stick-slip at the lubricant/solid interface and vice versa. Time needed for the transition differed depending on the separation between the opposing solid surfaces, between which lubricant molecules were filled. It was also shown that the delamination of lubricant layers on the solid surface was suppressed by the presence of solid surface asperity.

In recent advances in structural materials, a fundamental strategy for achieving strength and durability involves controlling microscopic heterogeneity, where different microstructures coexist within a single material. At the nanoscale level, the MAX phase, with general formula M_n+1AX_n (M = an early transition metal, A = an element typically from the 13th to the 16th group, X = C or N, and n = 1, 2, 3) exemplifies such a heterogeneous structure. It consists of a hard ceramic layer (M-X-M-X-M) and a relatively softer monatomic layer (M-A-M). The elastic state is also spatially inhomogeneous in such structures; however, a computational method to reveal this inhomogeneous elastic state at the atomic level has not yet been established. We have developed a local stiffness calculation scheme based on first-principles atomic stress calculations and applied it to the analysis of the MAX phase Ti₃AC₂ (A = Al, Ga, In, Si, Ge, Sn). The results confirmed the presence of elastic heterogeneity in the hard and soft layers. Specifically, it was observed that the soft layer stiffens when the element A moves from Period 5 to Period 3. For comparison, the elastic constants and the crystal orbital bonding index (COBI) of a single phase (Ti-A-Ti) were calculated to assess the influence of the interface and the electronic structure within the MAX phases. The elastic constants and ICOBI values in the soft layer within the MAX phase were generally larger than those in the single phase, clearly indicating the influence from the interface of the hard layer.

The nano-microstructure of polyamide is a layered structure of crystalline and amorphous phases called lamellae. Tie molecules that connect neighboring crystalline phases and entanglement points of molecular chains in the amorphous phase are called stress transmitters. They play an important role in macroscopic plastic deformation behavior. The behavior of stress transmitters during deformation has also been evaluated using molecular dynamics methods in previous studies. In this study, we prepared amorphous models for four types of PA (PA6, PA11, PAMXD10, and PA10T), and evaluated the effect of the difference in molecular chain morphology on stress increase during tensile deformation by all-atom molecular dynamics simulation. In addition, we investigate the entanglement of the molecular chains, the bond stretch and bending angle changes at each node, and discuss the relationship between the molecular chain morphology and the stress increase.

The stability of persistent slip band（PSB）at the submicron meter scale is investigated through molecular dynamics simulations. The PSB model includes grain boundaries in the direction of the dislocation line within the dislocation wall. Here, the special attention is paid to the “stability of the dislocation wall microstructure” and the “interaction between the screw dislocations moving in the channel and the grain boundary”. It is found that PSB cannot exist stably when the dimensions are small（for example, the channel distance is 60 nm）. Grain boundary is strong absorption site for channel screw dislocation, and once a screw dislocation is absorbed, it is difficult for it to be re-emitted. The operating stress and the re-emission stress of channel screw dislocation are greater than the collapse stresses of dislocation walls. As a result, the stability of PSB strongly depends on the plastic deformation ability of the screw dislocations in the channel in addition to the channel spacing. If the screw dislocations penetrate the grain boundary and are unable to relax plastically within the channel, the dislocation wall may collapse prematurely, suggesting that the formation of PSB in ultrafine-grained materials will be difficult.

As analytical models, bundle structures of carbon nanotubes (CNTs) were created by twisting together six (7, 7) CNTs. Simulation on compression behavior of the bundle structure was performed using molecular dynamics. As a result, it was found that the compressive strength decreases with an increase in the torsional angle of the bundle structure. We investigated the effect of the bundle structure ’s torsion on compressive deformation by evaluating the stress state of each individual CNT composing the bundle.

In this study, cyclic loading analysis of CNT/epoxy composite models were performed by numerical simulation using molecular dynamics. The composite models were subjected to 20 cycles of loading and unloading deformation in low strains. The analysis was performed at two temperatures, 300K and 1K. At both temperatures, the stress at unloading increased with the maximum strain due to non-elastic deformation. Differences in volume change after unloading were also observed depending on the direction of loading and temperature. Since there was no significant change in void volume, it is considered that the structural change by cyclic loading is mainly slight changes in bond distances and bond angles within the crosslinked structure.

Dislocation-precipitate interactions play an important role in determining the strength of alloys. Alloys can be strengthened by adding additive elements to the crystal and heat-treating to form precipitates, which hinder the movement of dislocations, suppress plastic deformation, and strengthen the metallic material. It is known that the coherency strain caused by the difference in lattice constants between the matrix and the precipitates affects the behavior of dislocations by producing a stress field around the precipitates. Various shapes of precipitates have been observed. In this study, a dislocation-precipitate interaction analysis considering the precipitate shape is performed using a hybrid computational method of the s-version finite element method and the dislocation dynamics method. The effect of precipitate shape on the critical resolved shear stress required for dislocations to shear the precipitate was investigated by calculating the stress field under coherency strain for spherical, disk, and cubic precipitate shapes.

Reactor pressure vessel (RPV) steels are known to be embrittled due to the neutron irradiations during the long-term operation of nuclear power plants. The neutron irradiation produces extensive amount of point defects, and finally forms matrix damages as a result of the self-diffusion of the point defects. As one of the matrix damage, prismatic dislocation loops are considered. Therefore, the interaction between prismatic dislocation loops and dislocations must be understood to clarify the role of the prismatic dislocation loop in the embrittlement of RPV steels. In this study, the interaction between prismatic dislocation loops and dislocations is investigated by the molecular dynamics (MD) method using an ANN potential to reveal the interaction mechanism and the critical resolved shear stress (CRSS). The numerical results show that, as a result of the direct interaction between the prismatic dislocation loop and a screw dislocation, a helical dislocation is formed. The increase in the CRSS due to the interaction with the prismatic dislocation loop is then modeled as an equation. It could be found that the prediction of the increase in the CRSS using the derived equation is quantitatively consistent with the MD simulation results.

To investigate the specificity of Σ15 grain boundaries in hetero-nanostructured Ti as a source of deformation modes of basal slip and {1121} twin, we perform tensile deformation simulations of a polycrystalline model containing random grain boundaries through molecular dynamics simulations. As a result, the following were found. (1) The basal dislocations are not activated even if it reached the CRSS of Σ15, but the prism dislocations are emitted from random grain boundaries preferentially. This suggests that Σ15 acts as an effective source of the basal dislocation. (2) The CRSS of the prism slip system emitted from Σ15 is much larger than the CRSS for random grain boundaries. This indicates that Σ15 is a difficult interface for prism dislocations to be activated. (3) The CRSS of {1011} twin activated at random grain boundaries was lower than the RSS of {1011} twin at Σ15. This indicates that Σ15 is a difficult interface for {1011} twin to be activated. In other words, it was recognized that {1121} twin was activated in Σ15 because {1121} twin reached the CRSS before {1011} twin.

Neutron irradiation enhances the formation of vacancies and Self Interstitial Atoms (SIAs), which cause the material embrittlement as a consequence of the interaction with dislocations. The vacancies and SIAs spontaneously form their clusters due to their self-diffusion. In this study, we calculated the activation energies and the minimum energy paths of five different diffusion mechanisms (Event 1 ~ 5) in α-iron by means of the Nudged Elastic Band (NEB) method with an Artificial Neural Network (ANN) interatomic potential. The result of the NEB analysis suggests that the activation energy of the Event 1 mechanism is the lowest and is 0.34 eV whereas an EAM potential gives the activation energy as 0.31 eV. The ANN potential can also reproduce well the activation energies and the minimum energy paths of the other mechanisms as well as the first principles calculations. The reaction pathway analysis also tells us that the Event 2 ~ 5 mechanisms can be reproduced by the multiple reptations of the Event 1 mechanism.

It is very important to determine the mobility of screw dislocation in body-centered cubic (BCC) iron. The size of dislocation core of screw dislocation is around 1 nm. Thus, atomic modeling is required to investigate the energetics of screw dislocation. Recently, using reference data based on density functional theory (DFT) calculations, we construct an artificial neural network (ANN) atomic potential to investigate the dislocation dynamics in BCC iron. The energetics of screw dislocation predicted by the constructed ANN potential are in good agreement with the reference DFT calculations. Using constructed ANN potential, in this study we investigate Peierls stress of the screw dislocation.

In this study, to develop a method to predict the ductile-brittle transition temperature (DBTT) of steels, a numerical analysis based on the fracture mechanism near the DBTT assuming microcrack cleavage in the vicinity of macrocrack was performed. In the model, the fracture behavior of microcracks and change in the fracture toughness with temperature are investigated by considering the temperature dependence of the motion of dislocations emitted from the crack-tip. The model shows that the activation of dislocation motion with increasing temperature strongly affects the shielding effect of the dislocations at the crack tip, resulting in a transition from brittle to non-brittle fracture, and the temperature related to the DBTT can be calculated. Furthermore, assuming the lattice defect-dislocation interaction, a friction stress was introduced to impede the motion of dislocations The results are qualitatively consistent with the actual behavior of the material.