Daegu Gyeongbuk Institute of Science and Technology
DGIST Library Institutional RepositoryNot a member yet
12664 research outputs found
Sort by
A Novel Compliance Compensator Capable of Measuring Six-Axis Force/Torque and Displacement for a Robotic Assembly
No full textThis article proposes a novel compliance compensator that can measure six-axis force, torque and displacement. The proposed device can provide the information necessary for feedback control while protecting the robot from impact. The device was designed based on a 6-DOF parallel mechanism to have six-axis sensing and deform greatly at the rated load without failure. We designed new flexure links with flexure joints connected in series, which facilitates fabrication and stiffness analysis. Through the stiffness analysis, the measured force can be converted to displacement and the stiffness of the device can be customized to the desired value by simply adjusting the thickness of the links. The sensing performance was evaluated through experiments using a commercial force/torque sensor (F/T sensor) and precision stages. We also propose a displacement-based misalignment compensation method in a robotic peg-in-hole assembly using the proposed device. The method uses the intrinsic passive compliance of the device and measured displacement, not the force and torque. We reduced the reaction force generated in the peg-in-hole assembly by 92.6% through the proposed method. The proposed method is simple and intuitive, and can be used for automatic teaching of the manipulator. © 2024 IEEEFALSEsciescopu
Throughput Approximation by Neural Network for Serial Production Lines With High Up/Downtime Variability
No full textMost of the existing studies on analyzing the productivity of serial production lines focus on cases where the coefficient of variation () for both uptime and downtime is less than 1. Hardly any result is available when , i.e., uptime and downtime of machines exhibit high variability. The improvement of the production lines with high variable uptime and downtime depends on heuristic trial and error due to the lack of analysis method. This article suggests a neural network that approximates the throughput of serial production lines from machine and buffer parameters. Four neural network architectures (multilayer perceptron, recurrent neural network, long short-term memory (LSTM), and gated recurrent unit) are compared to determine the most effective architecture for the throughput approximation task. Training data are obtained from discrete-event simulations, encompassing a wide range of parameters. The results indicate that the LSTM model outperforms the other architecture considered. Furthermore, we present bottleneck identification and continuous improvement scenarios utilizing the model. © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See https://www.ieee.org/publications/rights/index.html for more information.FALSEsciescopu
Magnetic anisotropy and magnetic phase diagram of a kagome antiferromagnet Fe1-xCoxSn
No full textWe investigate the anisotropic magnetic response in a kagome antiferromagnetic compound Fe1-xCoxSn (x = 0, 0.04, 0.06, 0.14). The magnetic behaviors of three distinct spin orderings, namely the planar, tilted, and axial spin states, are examined under various conditions, including doping rate, temperature, and magnetic fields. We confirm that the magnetic susceptibility of all three spin states has significant anisotropy, which depends on both the magnitude and direction of the applied magnetic field. The planar and axial states show scaling behavior in magnetic susceptibility, while the tilted state does not. Furthermore, employing the differential magnetic susceptibility, we observe that the axial spins display discontinuous spin-flop transitions, while the planar and tilted spins exhibit a continuous spin reorientation. The magnetic phase diagrams, summarizing our observations, highlight the anisotropic nature of magnetic behaviors in the Fe1-xCoxSn system. These diagrams clearly show that the magnetic responses are not overlapping and follow distinct mechanisms in their interaction with magnetic fields. We suggest that the axial state may arise due to the dopants as an obstacle within the kagome lattice, as evidenced by similar magnetic behaviors in Fe0.96Ni0.06Sn. © 2023 Elsevier B.V.FALSEsciescopu
Potential effects of HEK293 cell-derived exosomes for dermal application
No full text[No abstract available] © 2023 The Authors. Journal of Cosmetic Dermatology published by Wiley Periodicals LLC.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.TRUEsciescopu
Intelligent Channel Impulse Response Feature Prediction in Underwater Acoustic Networks
No full textIn this paper, we develop an intelligent channel im-pulse response (CIR) feature prediction algorithm in underwater networks. To this end, we first extract the major features, i.e., CIR values and tap distances, from raw CIR. Then, we implement a feature prediction module by adopting the time-series forecasting learning algorithm. Through the simulation results, we verify the prediction accuracy of the proposed algorithm with the normalized mean square error (NMSE) loss curve. © 2024 IEEE
CMAPSS Data Remaining Validity Prediction Using Deep Learning-Based Models with Parallel Connectivity Configuration
No full text터보팬 엔진의 RUL(Remaining useful life) 예측은 엔진이 고장 나기 전 예방 정비를 통해 엔진의 성능을 보장하고, 고장 전 엔진의 상태를 평가할 수 있다. 최근 많은 딥러닝 모델들이 RUL 예측을 시행하고 있으며 RUL 예측에 대한 좋은 성능을 보여주고 있다. 하지만 다소 복잡한 전처리 방법을 사용하여 RUL 예측을 위해 모델을 학습하는 경우가 많다. 이 방법은 비효율적인 노력이 소요될 수 있고, 모델의 복잡함에 의한 과적합을 초래할 수도 있다. 따라서 본 논문에서는 이러한 문제를 해결하기 위해 간단한 전처리 과정과 2차원 합성곱층(2D convolutional layer)을 병렬로 연결한 새로운 구조를 통해 문제를 해결할 수 있는 딥러닝 기반 방법을 제안한다. 먼저, 제안한 알고리즘의 구조에 대해서 설명하고 학습 데이터의 RUL 데이터 구성, 제안한 알고리즘에 대한 다양한 실험들과 RMSE 9.99로 기존 알고리즘들 성능 대비 0.5% 상승했음을 증명했다.
The prediction of the Remaining Useful Life (RUL) of a turbofan engine allows for evaluating the engine's condition before failure and ensuring performance through preventive maintenance. Recently, many deep learning models have been applied to RUL prediction, demonstrating good performance. However, many of these models rely on complex preprocessing methods for training, which can require inefficient effort and lead to overfitting due to model complexity. Therefore, this paper proposes a deep learning-based method to address these issues by using a simple preprocessing process and a novel structure that connects two-dimensional convolutional layers in parallel. Firstly, the structure of the proposed algorithm is explained, along with the configuration of the RUL data in the training dataset. Through various experiments conducted on the proposed algorithm, it demonstrated a 0.5% improvement over existing algorithms with an RMSE of 9.99
Simulation of autonomous system in severe weather
No full text자율주행 기술의 발전과 함께, 기상 조건이 시스템의 성능에 미치는 영향을 평가하는 것이 중요해지고 있으며, 다양한 악천후 상황은 센서의 정확도와 신뢰성에 큰 영향을 미칠 수 있다. 자율주행 시스템이 올바른 판단을 내리지 못할 경우, 운전자 및 동승자의 안전성과 직결된다. 본 논문에서는 시뮬레이션 환경을 통해 악천후 환경을 재현하여 자율주행 시스템의 센서 오작동을 소프트웨어 시뮬레이션하였다. 또한, 각 센서(예: 라이다, 카메라)의 반응과 성능을 확인하여 시스템의 안전성과 안전성을 향상시키기 위한 방법을 모색하도록 하였다. 본 논문은 자율주행 시스템이 다양한 기상 상황에서도 안전하게 작동하도록 방향성을 제시하고자 하였으며, 향후 실제 도로 주행 환경에서의 자율주행 기술의 신뢰성을 높이는 데 도움이 될 것으로 예상된다
Multi-material Surface Acoustic Wave Measuring Apparatus
No full text본 발명의 실시예에 따르면, 압전기판; 상기 압전기판에 형성되어 전기적 신호를 표면 탄성파 신호로 변환하는 송신부; 상기 송신부와 대향되는 위치에 형성되어 표면 탄성파 신호를 전기적 신호로 변환하는 복수 개의 센서부들이 1개의 송신부를 공유하도록 된 수신부;를 포함하고, 상기 센서부 또는 상기 센서부와 송신부 사이의 전파송신로 중 적어도 어느 하나에는 타겟 물질과 반응하는 활성물질이 도포되도록 하되, 상기 수신부는 송신부를 기점으로 방사상으로 펼쳐지는 원호를 따라서, 독립적으로 작동하는 적어도 2개 이상으로 이루어지는 복수 개의 센서부를 형성하는 다중물질 표면 탄성파 계측장치가 제공될 수 있다
MOBILE OTOSCOPE SYSTEM
No full text모바일 오토스코프 시스템이 개시된다. 본 모바일 오토스코프 시스템은 촬영 대상에 대한 이미지를 촬영하는 카메라를 포함하는 사용자 단말장치, 및 사용자 단말장치에 탈부착 가능하게 장착되며, 분광, UV 광 및 백색 광을 순차적으로 촬영 대상에 순차적으로 조사하는 모바일 오토스코프를 포함하고, 사용자 단말장치는 분광, UV 광 및 백색 광의 조사 중에 대상을 촬영하여 분광 이미지, UV 여기(excitation) 형광 이미지 및 3차원 형상 이미지를 순차적으로 생성하는 모바일 오토스코프 시스템
2D-3D Reconstruction of a Femur by Single X-Ray Image Based on Deep Transfer Learning Network
No full textObjective: Constructing a 3D model from its 2D images, known as 2D-3D reconstruction, is a challenging task. Conventionally, a parametric 3D model such as a statistical shape model (SSM) is deformed by matching the shapes in its 2D images through a series of processes, including calibration, 2D-3D registration, and optimization for nonrigid deformation. To overcome this complicated procedure, a streamlined 2D-3D reconstruction using a single X-ray image is developed in this study. Methods: We propose 2D-3D reconstruction of a femur by adopting a deep neural network, where the deformation parameters in the SSM determining the 3D shape of the femur are predicted from a single X-ray image using a deep transfer-learning network. For learning the network from distinct features representing the 3D shape information in the X-ray image, a specific proximal part of the femur from a unique X-ray pose that allows accurate prediction of the 3D femur shape is designated and used to train the network. Then, the corresponding proximal/distal 3D femur model is reconstructed from only the single X-ray image acquired at the designated position. Results: Experiments were conducted using actual X-ray images of a femur phantom and X-ray images of a patient's femur derived from computed tomography to verify the proposed method. The average errors of the reconstructed 3D shape of the proximal and distal femurs from the proposed method were 1.20 mm and 1.08 mm in terms of root mean squared point-to-surface distance, respectively. Conclusion: The proposed method presents an innovative approach to simplifying the 2D-3D reconstruction using deep neural networks that exhibits performance compatible with the existing methodologies. © 2024 AGBMFALSEsci