• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2020년도 하계학술대회
• /
• pp.540-543
• /
• 2020

기존의 초해상도 딥러닝 기법은 모델의 깊이가 깊어지면서, 좋은 성능을 내지만 점점 더 복잡해지고 있고, 실제로 사용하는데 있어 많은 시간을 요구한다. 이를 해결하기 위해, 우리는 딥러닝 모델의 가중치를 양자화 하여 추론시간을 줄이고자 한다. 초해상도 모델은 feature extraction, non-linear mapping, reconstruction 세 부분으로 나누어져 있으며, 레이어 사이에 많은 skip-connection 이 존재하는 특징이 있다. 따라서 양자화 시 최종 성능 하락에 미치는 영향력이 레이어 별로 다르며, 이를 감안하여 강화학습으로 레이어 별 최적 bit 를 찾아 성능 하락을 최소화한다. 본 논문에서는 Skip-connection 이 많이 존재하는 MSRN 을 사용하였으며, 결과에서 feature extraction, reconstruction 부분과 블록 내 특정 위치의 레이어가 항상 높은 bit 를 가짐을 알 수 있다. 기존에 영상 분류에 한정되어 사용되었던 혼합 bit 양자화를 사용하여 초해상도 딥러닝 기법의 모델 사이즈를 줄인 최초의 논문이며, 제안 방법은 모바일 등 제한된 환경에 적용 가능할 것으로 생각된다.

• Structural Engineering and Mechanics
• /
• 제6권3호
• /
• pp.259-274
• /
• 1998

The test results from non-destructive and destructive field testing of a three-span deteriorated reinforced concrete slab bridge are used as a vehicle to examine the reliability of available tools for finite-element analysis of in-situ structures. Issues related to geometric modeling of members and connections, material models, and failure criteria are discussed. The results indicate that current material models and failure criteria are adequate, although lack of inelastic out-of-plane shear response in most nonlinear shell elements is a major shortcoming that needs to be resolved. With proper geometric modeling, it is possible to adequately correlate the measured global, regional, and local responses at all limit states. However, modeling of less understood mechanisms, such as slab-abutment connections, may need to be finalized through a system identification technique. In absence of the experimental data necessary for this purpose, upper and lower bounds of only global responses can be computed reliably. The studies reaffirm that success of finite-element models has to be assessed collectively with reference to all responses and not just a few global measurements.

• 한국실내디자인학회논문집
• /
• 제13권2호
• /
• pp.29-37
• /
• 2004

This study analyze the architectural space in the aspect of the 'Raumplan' and 'Plan Libre'. These concepts based on the 'Raumplan versus Plan Libre' exhibition in Delft University, Netherlands. It is generally agreed that the concept of 'Plan Libre' in connection with Le Corbusie architecture and 'Raumplan' related to Adolf Loos's works. that exhibition contents was focused an extent of modem architecture. But, These concepts continually influence contemporary architecture and offer diverse vision in architecture. Therefore, this study focused on the analysis that look for common element in the space through the case study and space element connection in change of the values, technical growth. This Paper reveals that Adolf Loos's Raumplan is similar to Louis I. Kahn's 'Room' concept and Ando Tadao's centrifugal space composition method In the dwelling architecture. And the concept 'Plan Libre': non-definitive formal system and the elements : column, free form screening wall is revived as formal application of 3 dimension composition in architecture. and transfer lots, void, and linear space in Rem Koolhaas's urban design projects. These aspects is so called topology. This topological concept is an attempt in view of the space connection state rather than formative viewpoint.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2003년도 하계학술대회 논문집 D
• /
• pp.2633-2635
• /
• 2003

In this study, we propose a fuzzy polynomial neural networks (FPNN) and a genetically optimized fuzzy polynomial neural networks(GoFPNN) for identification of non-linear system. GoFPNN architecture is designed by a FPNN based on fuzzy set and its structure and parameters are optimized by genetic algorithms. A fuzzy neural networks(FNN) based on fuzzy set divide into two structures that is simplified inference structure and linear inference structure. The proposed FPNN is resulted from integration and extension of simplified and linear inference structure of FNN. The consequence structure of the FPNN consist of polynomials represented by networks using connection weights for rules. The networks comprehend simplified(Type 0), linear (Type 1), and quadratic(Type 3) inferences. The proposed FPNN can select polynomial type of consequence part for each rule. Therefore, proposed scheme can offer flexible structure design capability for a system characteristics. Moreover, GAs is applied to networks structure and parameters tuning of proposed FPNN, and its efficient application method is discussed, these subjects are result in GoFPNN that is optimal FPNN. To evaluate proposed model performance, a numerical experiment is carried out.

• 전기학회논문지
• /
• 제64권9호
• /
• pp.1337-1346
• /
• 2015

In this paper, pattern classifier is designed to classify precipitation and non-precipitation events from weather radar data. The proposed classifier is based on Fuzzy Neural Network(FNN) and consists of three FNNs which operate in parallel. In the proposed network, the connection weights of the consequent part of fuzzy rules are expressed as two polynomial types such as constant or linear polynomial function, and their coefficients are learned by using Least Square Estimation(LSE). In addition, parametric as well as structural factors of the proposed classifier are optimized through Differential Evolution(DE) algorithm. After event classification between precipitation and non-precipitation echo, non-precipitation event is to get rid of all echo, while precipitation event including non-precipitation echo is to get rid of non-precipitation echo by classifier that is also based on Fuzzy Neural Network. Weather radar data obtained from meteorological office is to analysis and discuss performance of the proposed event and echo patter classifier, result of echo pattern classifier compare to QC(Quality Control) data obtained from meteorological office.

• Structural Engineering and Mechanics
• /
• 제50권3호
• /
• pp.323-347
• /
• 2014

This paper proposes a Particle Swarm Optimization (PSO) algorithm, which is improved by making use of the Harmony Search (HS) approach and called HS-PSO algorithm. A computer code is developed for optimal sizing design of non-linear steel frames with various semi-rigid and rigid beam-to-column connections based on the HS-PSO algorithm. The developed code selects suitable sections for beams and columns, from a standard set of steel sections such as American Institute of Steel Construction (AISC) wide-flange W-shapes, such that the minimum total cost, which comprises total member plus connection costs, is obtained. Stress and displacement constraints of AISC-LRFD code together with the size constraints are imposed on the frame in the optimal design procedure. The nonlinear moment-rotation behavior of connections is modeled using the Frye-Morris polynomial model. Moreover, the P-${\Delta}$ effects of beam-column members are taken into account in the non-linear structural analysis. Three benchmark design examples with several types of connections are presented and the results are compared with those of standard PSO and of other researches as well. The comparison shows that the proposed HS-PSO algorithm performs better both than the PSO and the Big Bang-Big Crunch (BB-BC) methods.

• Structural Engineering and Mechanics
• /
• 제16권4호
• /
• pp.391-404
• /
• 2003

Local reinforcement in dowel type timber joints is essential to improve ductility, to increase load carrying capacity and to reduce the risk of brittle failure, especially in the case of using solid dowel. In many types of reinforcing materials available today, DVW (densified veneer wood) has been demonstrated to be the most advantages in terms of compatibility, embedding performance and ductility. Preliminary studies show that using appropriately sized DVW discs bonded into the timber interfaces may be an effective way to reinforce the connection. In this paper, non-linear 3-dimensional finite element models, incorporating orthotropic and non-linear material behaviour, have been developed to simulate structural performance of the timber joints locally reinforced by DVW discs. Different contact algorithms were applied to simulate contact conditions in the joints. The models were validated by the corresponding structural tests. Correlation between the experimental results and the finite element simulations is reasonably good. Using validated finite element models, parametric studies were undertaken to investigate effects of the DVW disc sizes and the end distances on shear stresses and normal stresses in a possible failure plane in the joint.

• 한국안전학회지
• /
• 제35권5호
• /
• pp.15-21
• /
• 2020

Quantitative risk assessment (QRA) is used as a legal or voluntary safety management tool for the hazardous material industry and the utilization of the method is gradually increasing. Therefore, a leak frequency analysis based on reliable generic data is a critical element in the evolution of QRA and safety technologies. The aim of this paper is to derive the leak frequency function that can be applied more flexibly in QRA based on OGP report with high reliability and global utilization. For the purpose, we first reviewed the data on the 16 equipments included in the OGP report and selected the predictors. And then we found good equations to fit the OGP data using non-linear regression analysis. The various expectation functions were applied to search for suitable parameter to serve as a meaningful reference in the future. The results of this analysis show that the best fitting parameter is found in the form of DNV function and connection function in natural logarithm. In conclusion, the average percentage error between the fitted and the original value is very small as 3 %, so the derived prediction function can be applicable in the quantitative frequency analysis. This study is to contribute to expand the applicability of QRA and advance safety engineering as providing the generic equations for practical leak frequency analysis.

• Advances in concrete construction
• /
• 제9권3호
• /
• pp.235-248
• /
• 2020

Progressive collapse in a structure occurs when load bearing members are failed and the adjoining structural elements cannot resist the redistributed forces and fails subsequently, that leads to complete collapse of structure. Recently, construction using precast concrete technology is adopted increasingly because it offers many advantages like faster construction, less requirement of skilled labours at site, reduced formwork and scaffolding, massive production with reduced amount of construction waste, better quality and better surface finishing as compared to conventional reinforced concrete construction. Connections are the critical elements for any precast structure, because in past, major collapse of precast structure took place because of connection failure. In this study, behavior of four different precast wet connections with U shaped reinforcement bars provided at different locations is evaluated. Reduced 1/3^rd scale precast beam column assemblies having two span beam and three columns with removed middle column are constructed and examined by performing experiments. The response of precast connections is compared with monolithic connection, under column removal scenario. The connection region of test specimens are filled by cast-in-place micro concrete with and without polypropylene fibers. Performance of specimen is evaluated on the basis of ultimate load carrying capacity, maximum deflection at the location of removed middle column, crack formation and failure propagation. Further, Finite element (FE) analysis is carried out for validation of experimental studies and understanding the performance of structural components. Monolithic and precast beam column assemblies are modeled using non-linear Finite Element (FE) analysis based software ABAQUS. Actual experimental conditions are simulated using appropriate boundary and loading conditions. Finite Element simulation results in terms of load versus deflection are compared with that of experimental study. The nonlinear FE analysis results shows good agreement with experimental results.

• 전기학회논문지
• /
• 제64권11호
• /
• pp.1543-1550
• /
• 2015

Ferroresonance is a non-linear vibrational phenomenon that is generated by the electrical interaction of the inductance component with the capacitor component of a certain capacitance as the device of the inductance component such as a transformer is saturated due to the degradation, the waveform distortion of current and voltage, and the oscillation of overcurrent and overvoltage in a system. Recently, ferroresonance was generated from the waveform distortion of current and voltage, or the overvoltage or undervoltage phenomenon caused by the nature of an electrical power system and design technology of the transformer in the three phase transformer system. Hence, in general, ferroresonance analyzed by converting to the LC equivalent circuit. However, in general, the aforementioned analytical method only applies to the resonance phenomenon that is generated by the interaction of the capacitance of bussbar and grounding, and switching as the capacitor component with PT and the transformer as the inductance component in a system. Subsequently, the condition where ferroresonance was generated since overvoltage was supplied as line voltage to the phase voltage and thus the iron core is saturated due to the interconnection between grounded and ungrounded systems could not be analyzed when single phase PT was connected in a ${\Delta}$/Y connection system. In this study, voltage swell in the configuration of grounded circuit of a step-up transformer with the ${\Delta}-{\Delta}$ connection linked to PT for control power and the ferroresonance generated by overvoltage when the line voltage of the ${\Delta}-{\Delta}$ connection was connected to the phase voltage of the grounded Y-Y connection were analyzed using PSCAD / EMTDC through the failure case of the transformer caused by ferroresonance in the system with the ${\Delta}-{\Delta}$/Y-Y connection, and subsequently, the preventive measure of ferroresonance was proposed.