• Journal of KIISE:Computer Systems and Theory
• /
• v.32 no.11_12
• /
• pp.682-691
• /
• 2005

We present a two-step evolution system that produces controllable virtual creatures in physically simulated 3D environment. Previous evolutionary methods for virtual creatures did not allow any user intervention during evolution process, because they generated a creature's shape, locomotion, and high-level behaviors such as target-following and obstacle avoidance simultaneously by one-time evolution process. In this work, we divide a single system into manageable two sub-systems, and this more likely allowsuser interaction. In the first stage, a body structure and low-level motor controllers of a creature for straight movement are generated by an evolutionary algorithm. Next, a high-level control to follow a given path is achieved by a neural network. The connection weights of the neural network are optimized by a genetic algorithm. The evolved controller could follow any given path fairly well. Moreover, users can choose or abort creatures according to their taste before the entire evolution process is finished. This paper also presents a new sinusoidal controller and a simplified hydrodynamics model for a capped-cylinder, which is the basic body primitive of a creature.

• Tunnel and Underground Space
• /
• v.29 no.3
• /
• pp.184-196
• /
• 2019

This study aims to extract a 3D image of micro-cracks generated by hydraulic fracturing tests, using the deep learning method and X-ray computed tomography images. The pixel-level cracks are difficult to be detected via conventional image processing methods, such as global thresholding, canny edge detection, and the region growing method. Thus, the convolutional neural network-based encoder-decoder network is adapted to extract and analyze the micro-crack quantitatively. The number of training data can be acquired by dividing, rotating, and flipping images and the optimum combination for the image augmentation method is verified. Application of the optimal image augmentation method shows enhanced performance for not only the validation dataset but also the test dataset. In addition, the influence of the original number of training data to the performance of the deep learning-based neural network is confirmed, and it leads to succeed the pixel-level crack detection.

• Proceedings of the Korean Institute of Surface Engineering Conference
• /
• 2012.11a
• /
• pp.123-123
• /
• 2012

최근 연구되고 있는 TSV(Through Silicon Via) 기술은 Si 웨이퍼 상에 직접 전기적 연결 통로인 관통홀을 형성하는 방법으로 칩간 연결거리를 최소화 할 수 있으며, 부피의 감소, 연결부 단축에 따른 빠른 신호 전달을 가능하게 한다. 이러한 TSV 기술은 최근의 초경량화와 고집적화로 대표되는 전자제품의 요구를 만족시킬 수 있는 차세대 실장법으로 기대를 모으고 있다. 한편, 납땜 재료의 주 원료인 주석은 주로 반도체 소자의 제조, 반도체 칩과 기판의 접합 및 플립 칩 (Flip Chip) 제조시의 범프 형성 등 반도체용 배선재료에 널리 사용되고 있다. 최근에는 납의 유해성 때문에 대부분의 전자제품은 무연솔더를 이용하여 제조되고 있지만, 주석을 이용한 반도체 소자가 고밀도화, 고 용량화 및 미세피치(Fine Pitch)화 되고 있기 때문에, 반도체 칩의 근방에 배치된 주석으로부터 많은 알파 방사선이 방출되어 메모리 셀의 정보를 유실시키는 소프트 에러 (Soft Error)가 발생되는 위험이 많아지고 있다. 이로 인해, 반도체 소자 및 납땜 재료의 주 원료인 주석의 고순도화가 요구되고 있으며, 특히 알파 방사선의 방출이 낮은 로우알파솔더 (Low Alpha Solder)가 요구되고 있다. 이에 따라 본 연구는 4인치 실리콘 웨이퍼상에 직경 $60{\mu}m$, 깊이 $120{\mu}m$의 비아홀을 형성하고, 비아 홀 내에 기능 박막증착 및 전해도금을 이용하여 전도성 물질인 Cu를 충전한 후 직경 $80{\mu}m$의 로우알파 Sn-1.0Ag-0.5Cu 솔더를 접합 한 후, 접합부 신뢰성 평가를 수행을 위해 고속 전단시험을 실시하였다. 비아 홀 내 미세구조와 범프의 형상 및 전단시험 후 파괴모드의 분석은 FE-SEM (Field Emission Scanning Electron Microscope)을 이용하여 관찰하였다. 연구 결과 비아의 입구 막힘이나 보이드(Void)와 같은 결함 없이 Cu를 충전하였으며, 고속전단의 경우는 전단 속도가 증가할수록 취성파괴가 증가하는 경향을 보였다. 본 연구를 통하여 전해도금을 이용한 비아 홀 내 Cu의 고속 충전 및 로우알파 솔더 볼의 범프 형성이 가능하였으며, 이로 인한 전자제품의 소프트에러의 감소가 기대된다.

• Journal of Biomedical Engineering Research
• /
• v.25 no.6
• /
• pp.497-503
• /
• 2004

Numerical simulation of blood flow has been conducted based on real vessel geometries generated front DICOM medical images of abdominal and iliac bifurcated arteries of a healthy man. A program was developed to read cross sectional images of the three dimensional arteries and smoothly extract boundary coordinates of vessels. Commercial programs were employed for mesh generation and flow simulation. Pressures, velocities, and flow distributions were found to lie within normal physiological ranges. Peak velocity measured in the iliac artery by ultrasound was 20% smaller than that obtained by simulation. The trend of velocity variation in a cardiac cycle was fairly similar between the simulation and the ultrasonic measurements. Simulation based on real vessel geometry of individual patient provides information on pressure, velocity, and its distribution in the diseased arteries or arteries to be surgically treated. The results of simulation may help surgeons to better understand hemodynamic status and surgical need of the patient by revealing variation of the hemodynamic parameters. Futhermore, they may serve as basic data for surgical treatment of arteries. This research is expected to develop to a program in the future that early diagnose atherosclerosis by showing distribution of a hemodynamic index closely related to atherosclerosis in arteries.