Due to the development of additive manufacturing (AM), lattice structure which cannot be fabricated by the conventional manufacturing process or have shape restriction has attracted much attention. ...We propose a new lightweight design method using two types of lattice structures considering the manufacturability in the metal selective laser melting (SLM) and structural characteristics. Firstly, the specific procedure for the proposed design method is presented. In order to apply the two lattice structures, relative density criterion is derived by fabricating experiments using metal SLM process and analyzing geometry according to relative density. The optimal relative density distribution is calculated by performing the topology optimization with minimum relative density using a commercial software package. This proposed method is computationally and experimentally validated by a three-point bending test. Simultaneously, the same procedure is applied to uniform lattice for comparison with the proposed method. This proposed design has a 46% increase in stiffness, a relative flexural rigidity of 35% compared to the solid material, and has a deformation mode different from the uniform lattice. This design sets the standard for using two lattice structures and gives a new perspective on lightweight design with lattice structures.
Display omitted
•The lattice structures fabricated by metal SLM process have excellent reproducibility.•The multi-lattice structures have superior stiffness and strength than a uniform lattice.•The area applied to the load is hardly deformed, and thus the sandwich core at center is well maintained.•The multi-lattice design can implement designs that control fracture behavior.
Federated learning, where the distribution of distributed data is unknown, is more difficult and costly to train a central model with than traditional machine learning. In this study, we propose ...Federated Learning with Genetic Algorithm, which enables faster central model training at lower cost by providing an appropriate client selection method. A client can have its own communication cost depending on its data sharing preference, and based on this cost and the result of the client's local update, we can select the appropriate combination of clients each round with a genetic algorithm. In each round, the client's combinations are evaluated anew, which are continually explored. To evaluate the algorithm, we distributed the image dataset and communication costs in two ways and conducted federated learning for the image classification model. Experiments showed that the proposed algorithm can find a more efficient client combination and accelerate the training of federated learning.
A novel red-emitting Na2Y2Ti3O10:Eu3+,Sm3+ phosphor was synthesized for the first time, and its photoluminescence (PL) properties were investigated for application to white-light-emitting diodes ...(W-LEDs). The Na2Y1Eu1Ti3O10 phosphor showed a 3-fold higher PL intensity compared to that of a commercial red-emitting Y2O2S:Eu3+ phosphor under 396 nm irradiation. Sm3+ ions were introduced as a sensitizer, and they increased the PL intensity by up to 35% under 410 nm irradiation. The effects of introducing the sensitizer were investigated by measuring the quantum efficiencies and the time-resolved PL decay curves. The Na2Y1Eu1Ti3O10 phosphor exhibited remarkable resistance to thermal quenching, and a W-LED comprising the phosphor showed bright white emission with a color rendering index (R a) of 83, a color temperature of 5556 K, and Commission Internationale de I’Eclairage (CIE) color coordinates of (0.3306, 0.3754).
The powder bed fusion (PBF) process using Ti–6Al–4V powder has the specific application in additive manufacturing of a high-performance structural parts in the aerospace and medical industries. The ...PBF involves the repeated accumulation of laser melted layers. Consequently, high anisotropic residual stresses and local temperature accumulation occur during the rapid melting and cooling in the process. These factors affect the mechanical properties of the as-built structure. In particular, we revealed the effective interval island laser-scanning strategy with less grain size, thermal effect and anisotropic residual stresses of the additively manufactured structure, compared to those of the strip and continuous laser-scanning strategies. Through the cantilever experiment, it was confirmed that the interval island laser-scanning strategy reduced deformation by up to 7.7% compared to that of the conventional strip laser-scanning strategy due to the reduction of anisotropic residual stresses.
In South Korea, where avian influenza virus subtypes H3N2, H5N1, H6N1, and H9N2 circulate or have been detected, 3 genetically similar canine influenza virus (H3N2) strains of avian origin ...(A/canine/Korea/01/2007, A/canine/Korea/02/2007, and A/canine/Korea/03/2007) were isolated from dogs exhibiting severe respiratory disease. To determine whether the novel canine influenza virus of avian origin was transmitted among dogs, we experimentally infected beagles with this influenza virus (H3N2) isolate. The beagles shed virus through nasal excretion, seroconverted, and became ill with severe necrotizing tracheobronchitis and bronchioalveolitis with accompanying clinical signs (e.g., high fever). Consistent with histologic observation of lung lesions, large amounts of avian influenza virus binding receptor (SAalpha 2,3-gal) were identified in canine tracheal, bronchial, and bronchiolar epithelial cells, which suggests potential for direct transmission of avian influenza virus (H3N2) from poultry to dogs. Our data provide evidence that dogs may play a role in interspecies transmission and spread of influenza virus.
Transfer printing is a materials assembly technique that uses elastomeric stamps for heterogeneous integration of various classes of micro‐ and nanostructured materials into two‐ and ...three‐dimensionally organized layouts on virtually any type of substrate. Work over the past decade demonstrates that the capabilities of this approach create opportunities for a wide range of device platforms, including component‐ and system‐level embodiments in unusual optoelectronic technologies with characteristics that cannot be replicated easily using conventional manufacturing or growth techniques. This review presents recent progress in functional materials and advanced transfer printing methods, with a focus on active components that emit, absorb, and/or transport light, ranging from solar cells to light‐emitting diodes, lasers, photodetectors, and integrated collections of these in functional systems, where the key ideas provide unique solutions that address limitations in performance and/or functionality associated with traditional technologies. High‐concentration photovoltaic modules based on multijunction, micro‐ and millimeter‐scale solar cells and high‐resolution emissive displays based on microscale inorganic light‐emitting diodes provide examples of some of the most sophisticated systems, geared toward commercialization.
Micro‐transfer printing is a materials assembly technique that enables heterogeneous integration of various classes of micro‐ and nanostructured materials. This article presents recent progress in functional materials and advanced transfer printing methods for active components that emit, absorb, and/or transport light, ranging from solar cells to light‐emitting diodes, lasers, photodetectors, and integrated functional systems, where the key ideas provide unique solutions to address limitations with traditional technologies.
The large amount of memory usage in recent machine learning applications imposes a significant system burden with respect to power and processing speed. To cope with such a problem, ...Processing-In-Memory (PIM) techniques can be applied as an alternative solution. Especially, the recommendation system, which is one of the major machine learning applications used in data centers, requires a large memory capacity and therefore represents a suitable candidate application that could be helped by the PIM technique. In this paper, we introduce a machine learning framework, PIMCaffe, designed for in-memory neural processing units and its evaluation environment. PIMCaffe consists of two components: a Caffe2-based deep learning framework that supports PIM acceleration and a PIM-emulating hardware platform. We develop a suite of functions, libraries, application programming interfaces, and a device driver to support the framework. In addition, we implement a prototype Neural Processing Unit (NPU) in PIMCaffe to evaluate the performance of our platform with machine learning applications. Our prototype NPU design includes a vector processor for parallel vector processing and a systolic array unit for matrix multiplication. Using the proposed software framework, we perform a detailed analysis of the in-memory neural processing unit. PIMCaffe supports evaluations of recommendation systems and various convolutional neural network models on the in-memory neural processing unit. PIMCaffe with the NPU shows up to <inline-formula> <tex-math notation="LaTeX">2.26\times </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">5.99\times </tex-math></inline-formula>, and <inline-formula> <tex-math notation="LaTeX">1.71\times </tex-math></inline-formula> speedup, compared to the ARM Cortex-A53 CPU, for the recommendation system, AlexNet, and ResNet-50, respectively.
Thermally engineered flexible composite assemblies of micro‐vertical cavity surface emitting lasers (VCSELs) are developed for high‐performance operation on a flexible substrate. The novel ...integration pathway enables significant reduction of the effective thermal resistance of printed micro‐VCSELs on plastics and therefore dramatically enhances performance, comparable to or higher than devices on the source wafer, while preserving critical advantages in mechanically flexible and optically transparent constructions.
Federated learning is a distributed learning algorithm designed to train a single server model on a server using different clients and their local data. To improve the performance of the server ...model, continuous communication with clients is required, and since the number of clients is very large, the algorithm must be designed in consideration of the cost required for communication. In this paper, we propose a method for distributing a model with a structure different from that of the server model, distributing a model suitable for clients with different data sizes, and training a server model using the reconstructed model trained by the client. In this way, the server model deploys only a subset of the sequential model, collects gradient updates, and selectively applies updates to the server model. This method of delivering the server model at a lower cost to clients who only need smaller models can reduce the communication cost of training server models compared to standard methods. An image classification model was designed to verify the effectiveness of the proposed method via three data distribution situations and two datasets, and it was confirmed that training was accomplished only with a cost 0.229 times smaller than the standard method.
PDF
