• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.934-936
• /
• 2020

기계학습 모델의 전체 구조를 쉽게 파악하고 추론할 수 있으며, 추론 과정 중에 멈춰서 중간결과를 확인할 수 있는 디버깅, 그리고 customizing 까지 지원하여 기계학습에 더 익숙해지고 더 나아가, 실제로 활용해보는 GUI Platform 구현

• Journal of the korean academy of Pediatric Dentistry
• /
• v.49 no.1
• /
• pp.85-94
• /
• 2022

This retrospective study aimed to evaluate the difference in measurement between conventional orthodontic analysis and artificial intelligence orthodontic analysis in pediatric and adolescent patients aged 7 - 15 with the mixed and permanent dentition. A total of 60 pediatric and adolescent patients (30 mixed dentition, 30 permanent dentition) who underwent lateral cephalometric radiograph for orthodontic diagnosis were randomly selected. Seventeen cephalometric landmarks were identified, and 22 measurements were calculated by 1 examiner, using both conventional analysis method and deep learning-based analysis method. Errors due to repeated measurements were assessed by Pearson's correlation coefficient. For the mixed dentition group and the permanent dentition group, respectively, a paired t-test was used to evaluate the difference between the 2 methods. The difference between the 2 methods for 8 measurements were statistically significant in mixed dentition group: APDI, SNA, SNB, Mandibular plane angle, LAFH (p < 0.001), Facial ratio (p = 0.001), U1 to SN (p = 0.012), and U1 to A-Pg (p = 0.021). In the permanent dentition group, 4 measurements showed a statistically significant difference between the 2 methods: ODI (p = 0.020), Wits appraisal (p = 0.025), Facial ratio (p = 0.026), and U1 to A-Pg (p = 0.001). Compared with the time-consuming conventional orthodontic analysis, the deep learning-based cephalometric system can be clinically acceptable in terms of reliability and validity. However, it is essential to understand the limitations of the deep learning-based programs for orthodontic analysis of pediatric and adolescent patients and use these programs with the proper assessment.

• Journal of Korean Society of Steel Construction
• /
• v.29 no.2
• /
• pp.111-122
• /
• 2017

In this paper, the applicability of a 1900mm-deep concrete-filled U-shaped composite beam to composite ordinary moment frames (C-OMFs) was investigated based on existing test results from smaller-sized specimens and supplemental numerical studies since full-scale seismic testing of such a huge sized beam is practically impossible. The key issue was the web local buckling of concrete-filled U section under negative bending. Based on 13 existing test results compiled, the relationship between web slenderness and story drift capacity was obtained. From this relationship, a 1900mm-deep mega beam, fabricated with 25mm-thick plate was expected to experience the web local buckling at 2% story drift and eventually reach a story drift over 3%, thus much exceeding the requirements of C-OMFs. The limiting width to thickness ratio according to the 2010 AISC Specification was shown to be conservative for U section webs of this study. The test-validated supplemental nonlinear finite element analysis was also conducted to further investigate the effects of the horizontal stiffeners (used to tie two webs of a U section) on web local buckling and flexural strength. First, it is shown that the nominal plastic moment under negative bending can be developed without using the horizontal stiffeners, although the presence of the stiffeners can delay the occurrence of web local buckling and restrain its propagation. Considering all these, it is concluded that the 1900mm-deep concrete-filled U-shaped composite beam investigated can be conservatively applied to C-OMFs. Finally, some useful recommendations for the arrangement and design of the horizontal stiffeners are also recommended based on the numerical results.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.8
• /
• pp.243-250
• /
• 2020

Knowledge graph-based question answering not only requires deep understanding of the given natural language questions, but it also needs effective reasoning to find the correct answers on a large knowledge graph. In this paper, we propose a deep neural network model for effective reasoning on a knowledge graph, which can find correct answers to complex questions requiring multi-hop inference. The proposed model makes use of highly expressive bilinear graph neural network (BGNN), which can utilize context information between a pair of neighboring nodes, as well as allows bidirectional feature propagation between each entity node and one of its neighboring nodes on a knowledge graph. Performing experiments with an open-domain knowledge base (Freebase) and two natural-language question answering benchmark datasets(WebQuestionsSP and MetaQA), we demonstrate the effectiveness and performance of the proposed model.

• Steel and Composite Structures
• /
• v.43 no.4
• /
• pp.447-456
• /
• 2022

Recent experimental studies showed that deep steel I-shaped profiles classified as high ductility class sections in seismic design international codes exhibit low deformation capacity when subjected to cyclic loading. This paper presents an innovative retrofit solution to increase the rotation capacity of beams using bonded carbon fiber reinforced polymers (CFRP) patches validated with advanced finite element analysis. This investigation focuses on the flexural cyclic behaviour of I-shaped hot rolled steel deep section used as beams in moment-resisting frames (MRF) retrofitted with CFRP patches on the web. The main goal of this CFRP reinforcement is to increase the rotation capacity of the member without increasing the overstrength in order to avoid compromising the strong column-weak beam condition in MRF. A finite element model that simulates the cyclic plasticity behavior of the steel and the damage in the adhesive layer is developed. The damage is modelled using the cohesive zone modelling (CZM) technique that is able to capture the crack initiation and propagation. Details on the modelling techniques including the mesh sensitivity near the fracture zone are presented. The effectiveness of the retrofit solution depends strongly on the selection of the appropriate adhesive. Different adhesive types are investigated where the CZM parameters are calibrated from high fidelity fracture mechanics tests that are thoroughly validated in the literature. This includes a rigid adhesive commonly found in the construction industry and two tough adhesives used in the automotive industry. The results revealed that the CFRP patch can increase the rotation capacity of a steel member considerably when using tough adhesives.

• International Journal of Concrete Structures and Materials
• /
• v.6 no.2
• /
• pp.65-78
• /
• 2012

An experimental study was carried out to evaluate fresh properties of a moderately high-strength (high-flowing) self-compacting concrete (SCC) and to investigate shear behavior and performance of deep beams made with SCC. Fresh and hardened properties of normal concrete (NC) and SCC were evaluated. The workability and compacting ability were observed based on casting time and number of surface cavities, respectively. Four-point loading tests on four deep beams (two made with SCC and two with NC) were then conducted to investigate their shear behavior and performance. Shear behavior and performance of beams having two different web reinforcements in shear were systematically investigated in terms of crack pattern, failure mode, and load-deflection response. It was found from the tests that the SCC specimen having a normal shear reinforcement condition exhibited a slightly higher load carrying capacity than the corresponding NC specimen, while the SCC specimen having congested shear reinforcement condition showed a similar load carrying capacity to the corresponding NC specimen. In addition, a comparative study between the present experimental results and theoretical results in accordance with ACI 318 (Building Code Requirements for Reinforced Concrete (ACI 318-89) and Commentary-ACI 318R-89, 1999), Hsu-Mau's explicit method (Hsu, Cem Concr Compos 20:419-435, 1998; Mau and Hsu, Struct J Am Concr Inst 86:516-523, 1989) and strut-and-tie model suggested by Uribe and Alcocer (2002) based on ACI 318 Appendix A (2008) was carried out to assess the applicability of the aforementioned methods to predict the shear strength of SCC specimens.

• Journal of Korean Society of Steel Construction
• /
• v.19 no.3
• /
• pp.247-258
• /
• 2007

This paper describes the results of an experimental study on the new type of encased composite beams that use deep deck plates, which could reduce the story height of buildings by controlling the bottom flange of steel beams. The profiled steel beam was thus developed. It was advantageous to the long span of the buildings. Seven full-scale specimens were constructed, and simply supported bending tests were conducted on the encased composite beams with different steel plate thicknesses, with and without shear studs, reinforcing bars, and web openings. The test results showed that the encased composite beams that were developed in this study had sufficient composite action without additional shear connectors due to their inherent shear-bond effects between the steel beams and concrete.

• Journal of Korea Society of Industrial Information Systems
• /
• v.27 no.5
• /
• pp.1-11
• /
• 2022

With the development of deep learning technology, various research and development are underway to estimate preference rankings through learning, and it is used in various fields such as web search, gene classification, recommendation system, and image search. Approximation algorithms are used to estimate deep learning-based preference ranking, which builds more than k comparison sets on all comparison targets to ensure proper accuracy, and how to build comparison sets affects learning. In this paper, we propose a k-disjoint comparison set generation algorithm and a k-chain comparison set generation algorithm, a novel algorithm for generating paired comparison sets for crowd-sourcing-based deep learning affinity measurements. In particular, the experiment confirmed that the k-chaining algorithm, like the conventional circular generation algorithm, also has a random nature that can support stable preference evaluation while ensuring connectivity between data.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.18 no.3
• /
• pp.768-776
• /
• 1994

This paper describes computer-aided forging design for rib/web shaped parts. In manufacturing a part by means of forging process, the first step is to design the forging. This is done by modifying the given machined part geometry according to the requirements of the forging process. Traditionally, this is done by experienced forging designers using empirical forging design guidelines. Generally, it would be neither possible nor practical to develop a system which encompasses the design of all types of forgings. Accordingly, forging design can be simplified by considering critical two dimensional cross sections of the machined part geometry. This system is composed of three modules(process variable decision module, forging design module and redesign module) and each module is carried out in regular sequence. In the process variable decision module, first of all, the undercut is checked and modified, and then deep recesses and holes difficult to forge are eliminated. Also parting line, forging plane, forging plan view area, forging weight and maximum size(maximum height or width)are determined. In the forging design module, the magnitude of various allowances, draft angle, minimum web thickness, corner and fillet radius are determined and then geometry modification is performed. Finally, since the design rules and databases used in this system are based on parameters of the forging geometry, such as the trimmed forging plan area, forging weight, forging maxmum size, plausible estimates need to be made for these parameters. Therefore, in the re-design module, the design process is iterated until a satisfactory forging is obtained.

• Steel and Composite Structures
• /
• v.35 no.3
• /
• pp.305-325
• /
• 2020

Beams of steel frame-tube structures (SFTSs) typically have span-to-depth ratios of less than five. This makes a flexural beam unsuitable for such an application because the plastic hinges at the beam-ends cannot be adequately developed. This leads to lower ductility and energy dissipation capacities of SFTSs. To address this, SFTSs with bolted web-connected replaceable shear links (SFTS-BWSLs) are proposed. In this structural system, a web-connected replaceable shear link with a back-to-back double channel section is placed at the mid-length of the deep beam to act as a ductile fuse. This allows energy from earthquakes to be dissipated through link shear deformation. SFTS and SFTS-BWSL buildings were examined in this study. Several sub-structures were selected from each designed building and finite element models were established to study their respective hysteretic performance. The seismic behavior of each designed building was observed through static and dynamic analyses. The results indicate that the SFTS-BWSL and SFTS have similar initial lateral stiffness and shear leg properties. The SFTS-BWSL had lower strength, but higher ductility and energy dissipation capacities. Compared to the SFTS, the SFTS-BWSL had lower interstory drift, base shear force, and story shear force during earthquakes. This design approach could concentrate plasticity on the shear link while maintaining the residual interstory drift at less than 0.5%. The SFTS-BWSL is a reliable resistant system that can be repaired by replacing shear links damaged due to earthquakes.