• Korean Journal of Computational Design and Engineering
• /
• v.6 no.2
• /
• pp.111-124
• /
• 2001

In this paper, we propose rounding operations on shell meshes, which give a constant or variable radius of rounding directly to sharp edges on a shell mesh. The rolling-ball methods for freeform surface rounding are introduced to devise the algorithms for these operations. Our algorithms consists of three steps as follows: detecting sharp edges, generating a rolling-ball surface contacting with two face groups adjacent to the sharp edges, and then replacing the rounding area of the original mesh with the mesh generated on the rolling-ball surface. In addition, this paper shows their application to the area of stamping die design. These operations enable CAE engineers to directly change the meshes of stamping tools without modification of CAD models for dies and regeneration of their meshes.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.638-643
• /
• 2023

문서 기반 대화 시스템은 크게 질문으로부터 문서를 검색하는 과정과 응답 텍스트를 생성하는 과정으로 나뉜다. 이러한 대화 시스템의 응답 생성 과정에 디코더 기반 LLM을 사용하기 위해서 사전 학습된 LLM을 미세 조정한다면 많은 메모리, 연산 자원이 소모된다. 본 연구에서는 SVD에 기반한 LLM의 경량화를 시도한다. 사전 학습된 polyglot-ko 모델의 행렬을 SVD로 분해한 뒤, full-fine-tuning 해보고, LoRA를 붙여서 미세 조정 해본 뒤, 원본 모델을 미세 조정한 것과 점수를 비교하고, 정성평가를 수행하여 경량화된 모델의 응답 생성 성능을 평가한다. 문서 기반 대화를 위한 한국어 대화 데이터셋인 KoDoc2Dial에 대하여 평가한다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.399-403
• /
• 2023

본 논문에서는 다중 문서 기반의 대화 시스템을 통한 효율적인 정보 검색과 응답 생성에 중점을 둡니다. 대규모 데이터 집합에서 정확한 문서를 선택하는 데 필요한 검색의 중요성을 강조하며, 현재 검색 방법의 한계와 문제점을 지적합니다. 또한 더 자연스러운 답변을 생성하기 위해 대규모 언어 모델을 사용하게 되면서 fine-tuning 시에 발생하는 제약과 낭비를 모델의 제로샷 생성 능력을 활용하여 개선하려는 방안을 제안하며, 모델의 크기와 자원의 효율성에 대한 고려사항을 논의합니다. 우리의 접근 방식은 대규모 언어 모델을 프롬프트와 함께 다중 문서로 학습 없이 정보를 검색하고 응답을 생성하는 방향으로 접근하여 대화 시스템의 효율성과 유용성을 향상시킬 수 있음을 제시합니다.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.37 no.7
• /
• pp.865-873
• /
• 2013

The deep drawing and ironing (DDI) process involving the use of a high-capacity horizontal press is used for manufacturing acompressed natural gas (CNG) pressure vessel. However, some variables of the DDI process have been determined based on the experiences of workers, and the short die life needs to be improved for manufacturing the pressure vessel withhighquality and lowcost. In this study, process variables such as the draw ratio, distance between dies, radius of rounding of drawing die, and angle of ironing die are chosen to enhance the reliability and improve the die life based on previous studies and experiences. The draw ratio limits at which no tearing or wrinkling occurs are determined using FEA, and the distances between dies, radius of rounding of drawing die, and angle of ironing die are optimized by the DOE method. The results of the optimal process variables are compared with those of the existing DDI process for verifying their effectiveness.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.48 no.10
• /
• pp.72-81
• /
• 2011

In this paper, we proposed and designed a novel floating point converter which supports single and double precisions of IEEE754 standard. The proposed convertor supports conversions between floating point number single/double precision and signed fixed point number(32bits/64bits) as well as conversions between signed integer(32bits/64bits) and floating point number single/double precision and conversions between floating point number single and double precisions. We defined a new internal format to convert various input types into one type so that overflow checking could be conducted easily according to range of output types. The internal format is similar to the extended format of floating point double precision defined in IEEE754 2008 standard. This standard specifies that minimum exponent bit-width of the extended format of floating point double precision is 15bits, but 11bits are enough to implement the proposed converting unit. Also, we optimized rounding stage of the convertor unit so that we could make it possible to operate rounding and represent correct negative numbers using an incrementer instead an adder. We designed single cycle data path and 5 cycles data path. After describing the HDL model for two data paths of the convertor, we synthesized them with TSMC 180nm technology library using Synopsys design compiler. Cell area of synthesis result occupies 12,886 gates(2 input NAND gate), and maximum operating frequency is 411MHz.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.9A
• /
• pp.1452-1460
• /
• 2000

A series of papers in recent literature on logical topology design for wavelength routed optical networks have proposed mathematical models and solution methods unifying logical topology design wavelength assignment and traffic routing. The most recent one is by Krishnaswamy and Sivarajan which is more unifying and complete than the previous models. Especially the mathematical formulation is an integer linear program and hence regarded in readiness for an efficient solution method compared to the previous nonlinear programming models. The solution method in [7] is however elementary one relying on the rounding of linear program relaxation. When the rounding happens to be successful it tends to produce near-optimal solutions. In general there is no such guarantee so that the obtained solution may not satisfy the essential constraints such as logical -path hop-count and even wavelength number constraints. Also the computational efforts for linear program relaxation seems to be too excessive. In this paper we propose an improved and unifying solution method based on the same to be too excessive. In this paper we propose an improved and unifying solution method based on the same model. First its computation is considerably smaller. Second it guarantees the solution satisfies all the constraints. Finally applied the same instances the quality of solution is fairly competitive to the previous near optimal solution.

• Journal of Biomedical Engineering Research
• /
• v.21 no.6
• /
• pp.599-606
• /
• 2000

고정체 나사산 형상이 식립된 고정체를 둘러싸고 있는 턱 뼈에서 발생된 응력분포에 미치는 영향과 효과적인 나사산 형상을 결정하기 위해서 다양한 치아 고정체 형상에 대해서 응력해석을 수행하였다. 나사산 골 한쪽 부위에 라운딩이 된 형상의 고정체에서 발생된 응력분포는 다른 나사산 형상의 고정체에서 발생된 응력분포보다 더 효과적으로 나타났다. 이 해석 결과를 근거로 최적의 고정체 치수를 결정하기 위해서 나사산 끝단의 폭, 나사산 높이, 그리고 가해지는 하중의 방향 등과 같은 설계 변수의 변화에 따른 응력해석이 수행되었다. 최대 응력 집중은 고정체 나사산의 첫단 부위에서 발생하였으며, 100 N의 15도 경사하중이 가해졌을 때 발생된 최대 등가응력은 동일 크기의 수식하중보다 2배 정도 더 높게 나타났다. 그리고 나사산 끝단의 폭과 나사산 높이 사이의 연관성에 관련된 해석결과에서 나사산 끝단의 폭과 나사산 높이 사이의 연관성 효과는 무시할 만큼 작다는 것을 알았다. 고정체의 나사산 피치에 대한 나사산 끝단의 폭의 비와 나사산 높이의 비가 각각 0.5와 0.46일 때 다른 고정체의 치수들보다 더 효과적인 응력분포가 나타났다

• Journal of the Korean Society of Radiology
• /
• v.9 no.2
• /
• pp.73-77
• /
• 2015

In order to observe the detailed structure of the detector, it was CT scanned to reproduce the detailed structure of the crystal shapes and traverse layer using the Monte Carlo calculation applying the detector model. The uncertainty of measurement was lowered by adjusting the detector core by the edge effect at a higher energy (400 keV or higher) through the offset of peak efficiency of the gamma ray at low energy. It was confirmed that there was the appropriate matching with spatial dependency using the PENELOPE calculation. That was achieved by adjusting the parameters describing the crystal core and rounding of edge and crystal core.

• Journal of the Korean Society of Industry Convergence
• /
• v.25 no.3
• /
• pp.407-415
• /
• 2022

In this study, a square rounding sidewalk block frame was developed considering the simplicity of construction and the superiority of aesthetics. In addition, it is possible to prevent damage, deformation, and settlement of adjacent sidewalk blocks due to plastic deformation during car impact load of installed bollad. A non-linear structural analysis was performed through finite element analysis to examine the performance of a car crash to which this was applied. Structural safety was confirmed through car crash simulation according to the direction of impact, and it is estimated that the function can be restored by replacing some parts in case of damage due to impact.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.16 no.5
• /
• pp.66-74
• /
• 2002

OFDM(Orthogonal Frequency Division Multiplexing)is a one of the nst promising digital modulation techniques adapted for Digital audio broadcasting or Digital TV since it is very robust against multipath fading channels. From 1997, since the OFDM technique was considered as the physical layer standard for the high data rate wireless LAN systems in the 5㎓ band, related studies have been studied actively. The key element to implement high data rate wireless LAN system using OFDM technique are IFFT and FFT modules. In this paper, new IFFT and FFT module are designed and implemented using current cut circuit based on the matrix-rounding process for the low-power consumptive operation and high-speed data processing. In addition to, we certify the available operation of the rounded IFFT/FFT module in the AWGN channel by using the BER performance simulation of IEEE 802.11TGa based OFDM modem with rounded IFFT/FFT module.