• Earthquakes and Structures
• /
• 제7권3호
• /
• pp.333-347
• /
• 2014

Multi-scale model can take both computational efficiency and accuracy into consideration when it is used to conduct elasto-plastic seismic response analysis for complex steel bridges. This paper proposed a method based on pushover analysis of member sharing the same section pattern to verify the accuracy of multi-scale model. A deck-through type steel arch bridge with a span length of 200m was employed for seismic response analysis using multi-scale model and fiber model respectively, the validity and necessity of elasto-plastic seismic analysis for steel bridge by multi-scale model was then verified. The results show that the convergence of load-displacement curves obtained from pushover analysis for members having the same section pattern can be used as a proof of the accuracy of multi-scale model. It is noted that the computational precision of multi-scale model can be guaranteed when length of shell element segment is 1.40 times longer than the width of section where was in compression status. Fiber model can only be used for the predictions of the global deformations and the approximate positions of plastic areas on steel structures. However, it cannot give exact prediction on the distribution of plastic areas and the degree of the plasticity.

• East Asian mathematical journal
• /
• 제33권1호
• /
• pp.23-36
• /
• 2017

We study an optimization problem for HARA utility function under a multi-scale Heston's stochastic volatility model. We investigate a practical strategy that do not depend on the incorporated factor which is unobservable in the market.

• Journal of Multimedia Information System
• /
• 제8권4호
• /
• pp.203-210
• /
• 2021

Deep leaning convolutional neural networks (CNN) have successfully been applied to image super-resolution (SR). Despite their great performances, SR techniques tend to focus on a certain upscale factor when training a particular model. Algorithms for single model multi-scale networks can easily be constructed if images are upscaled prior to input, but sub-pixel convolution upsampling works differently for each scale factor. Recent SR methods employ multi-scale and multi-path learning as a solution. However, this causes unshared parameters and unbalanced parameter distribution across various scale factors. We present a multi-scale single-path upsample module as a solution by exploiting the advantages of sub-pixel convolution and interpolation algorithms. The proposed model employs sub-pixel convolution for the highest scale factor among the learning upscale factors, and then utilize 1-dimension interpolation, compressing the learned features on the channel axis to match the desired output image size. Experiments are performed for the single-path upsample module, and compared to the multi-path upsample module. Based on the experimental results, the proposed algorithm reduces the upsample module's parameters by 24% and presents slightly to better performance compared to the previous algorithm.

• Coupled systems mechanics
• /
• 제11권1호
• /
• pp.55-78
• /
• 2022

The main aim of this paper is the identification of the model parameters for the constitutive model of concrete and concrete-like materials capable of representing full set of 3D failure mechanisms under various stress states. Identification procedure is performed taking into account multi-scale character of concrete as a structural material. In that sense, macro-scale model is used as a model on which the identification procedure is based, while multi-scale model which assume strong coupling between coarse and fine scale is used for numerical simulation of experimental results. Since concrete possess a few clearly distinguished phases in process of deformation until failure, macro-scale model contains practically all important ingredients to include both bulk dissipation and surface dissipation. On the other side, multi-scale model consisted of an assembly micro-scale elements perfectly fitted into macro-scale elements domain describes localized failure through the implementation of embedded strong discontinuity. This corresponds to surface dissipation in macro-scale model which is described by practically the same approach. Identification procedure is divided into three completely separate stages to utilize the fact that all material parameters of macro-scale model have clear physical interpretation. In this way, computational cost is significantly reduced as solving three simpler identification steps in a batch form is much more efficient than the dealing with the full-scale problem. Since complexity of identification procedure primarily depends on the choice of either experimental or numerical setup, several numerical examples capable of representing both homogeneous and heterogeneous stress state are performed to illustrate performance of the proposed methodology.

• 한국멀티미디어학회논문지
• /
• 제21권12호
• /
• pp.1467-1472
• /
• 2018

For the problem that the face image of surveillance video cannot be accurately identified due to the low resolution, this paper proposes a low resolution face recognition solution based on convolutional neural network model. Convolutional Neural Networks (CNN) model for multi-scale input The CNN model for multi-scale input is an improvement over the existing "two-step method" in which low-resolution images are up-sampled using a simple bi-cubic interpolation method. Then, the up sampled image and the high-resolution image are mixed as a model training sample. The CNN model learns the common feature space of the high- and low-resolution images, and then measures the feature similarity through the cosine distance. Finally, the recognition result is given. The experiments on the CMU PIE and Extended Yale B datasets show that the accuracy of the model is better than other comparison methods. Compared with the CMDA_BGE algorithm with the highest recognition rate, the accuracy rate is 2.5%~9.9%.

• 한국자동차공학회논문집
• /
• 제7권7호
• /
• pp.1-14
• /
• 1999

In the present study, the flame factor which primarily influence the simulation accuracy of the combustion process in a gasoline engine was modeled as a nonlinear function of turbulent intensity to laminar flame speed ratio. Multi-length-scale production rate model for turbulent kinetic energy equation was introduced to consider the different length scales of the swirling and tumbling motions in cylinder on the production rte of turbulent kinetic energy. By7 introducing the multi-length-scale production rate model for the turbulent kinetic energy equation, the predictions of turbulent burning velocity , cylinder pressure, mass burning rate and engine performance of a gasoline engine can much be improved.

• Computers and Concrete
• /
• 제31권1호
• /
• pp.9-21
• /
• 2023

The published studies usually used analytical method, numerical methods or experimental method to determine the stress-strain state and displacement of the single-layer or multi-layer curved shell types, but with a small scale model. However, a full scale multi-layer doubly curved concrete shell roof model should be researched. This paper presents the results of the experiment and simulation analysis involving stress-strain state, sliding between layers, the formation and development of the full scale double-layer doubly curved concrete shell roof when this shell begins to crack. The results of the this study have constructed the load-sliding strain relationship; strain diagram; stress diagram in the shell layers; the N_x, N_y membrane force diagram and deflection of shell. Thisresults by experimental method on a full scale model of concrete have clarified the working of multi-layer doubly curved concrete shell roof. The experimental and simulation results are compared with each other and compared with the Sap2000 software.

• 제어로봇시스템학회논문지
• /
• 제10권12호
• /
• pp.1164-1171
• /
• 2004

A new multi-scale simulation model is proposed to analyze heart mechanics. Electrophysiology of a cardiac cell is numerically approximated using the previous model of human ventricular myocyte. The ion transports across cell membrane initiated by action potential induce an excitation-contraction mechanism in the cell via cross bridge dynamics. Negroni and Lascano model (NL model) is employed to calculate the tension of cross bridge which is closely related to the ion dynamics in cytoplasm. To convert the tension on cell level into contraction force of cardiac muscle, we introduce a simple geometric model of ventricle with a thin-walled hemispheric shape. It is assumed that cardiac tissue is composed of a set of cardiac myocytes and its orientation on the hemispheric surface of ventricle remains constant everywhere in the domain. Application of Laplace law to the ventricle model enables us to determine the ventricular pressure that induces blood circulation in a body. A lumped parameter model with 7 compartments is utilized to describe the systemic circulation interacting with the cardiac cell mechanism via NL model and Laplace law. Numerical simulation shows that the ion transports in cell level eventually generate blood hemodynamics on system level via cross bridge dynamics and Laplace law. Computational results using the present multi-scale model are well compared with the existing ones. Especially it is shown that the typical characteristics of heart mechanics, such as pressure volume relation, stroke volume and ejection fraction, can be generated by the present multi-scale cardiovascular model, covering from cardiac cells to circulation system.

• Nuclear Engineering and Technology
• /
• 제55권3호
• /
• pp.1140-1151
• /
• 2023

Fully Ceramic Microencapsulated (FCM) fuel is emerging advanced fuel material for the future nuclear reactors. The fuel pellet in the FCM fuel is composed of matrix and a large number of TRistructural-ISOtopic (TRISO) fuel particles which are randomly dispersed in the SiC matrix. The minimum layer thickness in a TRISO fuel particle is on the order of 10^-5 m, and the length of the FCM pellet is on the order of 10^-2 m. Hence, the heat transfer in the FCM pellet is a multi-scale phenomenon. In this study, three multi-scale heat conduction models including the Multi-region Layered (ML) model, Multi-region Non-layered (MN) model and Homogeneous model for FCM pellet were constructed. In the ML model, the random distributed TRISO fuel particles and coating layers are completely built. While the TRISO fuel particles with coating layers are homogenized in the MN model and the whole fuel pellet is taken as the homogenous material in the Homogeneous model. Taking the results by the ML model as the benchmark, the abilities of the MN model and Homogenous model to predict the maximum and average temperature were discussed. It was found that the MN model and the Homogenous model greatly underestimate the temperature of TRISO fuel particles. The reason is mainly that the conventional equivalent thermal conductivity (ETC) models do not take the internal heat source into account and are not suitable for the TRISO fuel particle. Then the improved ETCs considering internal heat source were derived. With the improved ETCs, the MN model is able to capture the peak temperature as well as the average temperature at a wide range of the linear powers (165 W/cm~ 415 W/cm) and the packing fractions (20%-50%). With the improved ETCs, the Homogenous model is better to predict the average temperature at different linear powers and packing fractions, and able to predict the peak temperature at high packing fractions (45%-50%).

• Journal of Information Processing Systems
• /
• 제19권4호
• /
• pp.427-438
• /
• 2023

Plenty of works have indicated that single image super-resolution (SISR) models relying on synthetic datasets are difficult to be applied to real scene text image super-resolution (STISR) for its more complex degradation. The up-to-date dataset for realistic STISR is called TextZoom, while the current methods trained on this dataset have not considered the effect of multi-scale features of text images. In this paper, a multi-scale and attention fusion model for realistic STISR is proposed. The multi-scale learning mechanism is introduced to acquire sophisticated feature representations of text images; The spatial and channel attentions are introduced to capture the local information and inter-channel interaction information of text images; At last, this paper designs a multi-scale residual attention module by skillfully fusing multi-scale learning and attention mechanisms. The experiments on TextZoom demonstrate that the model proposed increases scene text recognition's (ASTER) average recognition accuracy by 1.2% compared to text super-resolution network.