• 수산해양기술연구
• /
• 제59권2호
• /
• pp.99-109
• /
• 2023

In this study, a comparative test operation was conducted through the alternate haul method to examine the selectivity of the four mesh sizes (60 mm, 90 mm, 110 mm, and 130 mm) of the trawl codend. The selectivity was analyzed using the SELECT model considering the fishing efficiency (split parameter) of each fishing gear in the comparative test fishing operation in the trawl and the maximum likelihood method for parameter estimation. A selectivity master curve was estimated for several mesh sizes using the extended-SELECT model. As a result of analyzing the selectivity for silver croaker based on the results of three times hauls for each experimental gear, it was found that the size of the fish caught increased as the size of the mesh size increased. When the selectivity for each mesh size analyzed by the SELECT model considering the split ratio was evaluated based on the size of the AIC value, the estimated split model was superior to the equal split model. Based on the master curve, the 50% selection length value was 2.893, which was estimated to be 136 mm based on the mesh size of 60 mm. In some selectivity models, there was a large deviance between observed and theoretical values due to the non-uniformity of the distribution of fished length classes. As a result, it is considered that appropriate sea trials and selectivity evaluation methods with high reliability should be applied to present trawl fishery resource management methods.

• 지구물리와물리탐사
• /
• 제25권4호
• /
• pp.227-241
• /
• 2022

전파형 역산은 석유가스 탐사를 위한 탄성파 자료처리 분야에서 지층의 속도 모델을 추정하는데 사용되는 역산 기법이다. 최근 탄성파 자료처리에 딥러닝 기술의 활용이 급격하게 증가하고 있는데, 전파형 역산 기술도 마찬가지로 다양한 연구가 이루어지고 있다. 초기에는 머신러닝 기술을 활용한 자료처리 기법이 전파형 역산을 위한 입력자료의 전처리 목적으로 활용되는 수준이었으나, 딥러닝 기술을 통해 전파형 역산을 직접적으로 구현하는 연구가 등장하기 시작하였다. 딥러닝 기술을 활용한 전파형 역산은 순수 데이터 기반 접근법, 물리 기반 신경망 활용법, 인코더-디코더 구조 활용법, 신경망 재매개변수화를 이용한 구현법, 물리정보 기반 신경망 기법 등으로 구분할 수 있다. 이 논문에서는 딥러닝 기반 전파형 역산 기법을 발전 과정 순서로 체계화하여 각각의 접근법에 대한 이론과 특징을 설명하였다. 전파형 역산 기술에 딥러닝 기법을 도입한 초기에는 데이터 과학의 기본 원리에 충실하게 대량의 학습자료를 준비하고 순수 데이터 기반 예측 모델을 적용하여 속도 모델을 역산하는 연구로 시작하였다. 최근 연구 동향은 탄성파 자료의 잔차나 파동방정식 자체의 물리정보를 심층 신경망에 활용하여 순수 데이터 기반 접근법의 단점을 보완해 나가는 방향으로 진행되고 있다. 이러한 발전으로 대량의 학습자료가 필요하지 않고, 전파형 역산의 태생적 한계점인 주기 놓침 현상을 완화하며 계산 시간을 획기적으로 줄일 수 있는 딥러닝 기반 전파형 역산 기술이 등장하고 있다. 딥러닝 기술의 도입으로 전파형 역산 기술은 탄성파 자료처리 분야에서 가치가 더 높아질 것으로 생각된다.

• EDISON SW 활용 경진대회 논문집
• /
• 제5회(2016년)
• /
• pp.368-372
• /
• 2016

In this paper, the physics-based analytical model of transition metal dichalcogenide (TMD)-based double-gate (DG) tunneling field-effect transistors (TFETs) is proposed. The proposed model is derived by using the two-dimensional (2-D) Landauer formula and the Wentzel-Kramers-Brillouin (WKB) approximation. For improving the accuracy, nonlinear and continuous lateral energy band profile is applied to the model. 2-D density of states (DOS) and two-band effective Hamiltonian for TMD materials are also used in order to consider the 2-D nature of TMD-based TFETs. The model is validated by using the tight-binding non-equilibrium Green's function (NEGF)-based quantum transport simulation in the case of monolayer molybdenum disulfide ($MoS_2$)-based TFETs.

• 천문학회보
• /
• 제40권1호
• /
• pp.46.2-46.2
• /
• 2015

Modern methods in determining the value of the Hubble constant are divided into two main ways: the classical distance ladder method and the inverse distance ladder method. The classical distance ladder method is based on Cepheid calibrated Type Ia supernovae (SNe Ia), which are known as powerful distance indicator. The inverse distance ladder method uses cosmic microwave background radiation, which emitted from the high-z universe, and the cosmological model. Recent estimations of the Hubble constant based on these two methods show a $2{\sim}3{\sigma}$ difference, which called the "Hubble tension". It is currently an issue in the modern cosmology. We have been working on the luminosity calibration of SNe Ia based on the Tip of the Red Giant Branch (TRGB), which is a precise population I distance indicator. We present the TRGB distance estimates of 5 SNe Ia host galaxies with the archival Hubble Space Telescope image data. We derive the mean absolute maximum magnitude of 5 SNe Ia and the value of the Hubble constant. Cosmological implications of our estimate will be discussed.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 2008년도 추계학술대회 논문집
• /
• pp.288-291
• /
• 2008

Physics based Cellular Automata model is developed and implemented into FEM code. CA model can predict microstructure evolution based on physical phenomena, such as hardening, recovery and recrystallization. This paper outlines the methodology to determine the materials constants for these different phenomena from simpler measurements.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.1115-1117
• /
• 2005

This paper describes the three dimension model and simulation results of a thermal actuator based on polyMUMPs process, known as thermal multimorph actuator. The device has potential application in micro-transducers such as atomic force microscope (AFM) tip and scanning tunneling microscope (STM) tip. This device made of a multi-layer materials stack together with consisted of polysilicon, $SiO_2$ and gold. A mask layout design, three dimension model and simulation results are reported and discussed.

• Nuclear Engineering and Technology
• /
• 제49권6호
• /
• pp.1219-1225
• /
• 2017

In this work, a scalable algorithm for model calibration in nuclear engineering applications is presented and tested. The algorithm relies on the construction of surrogate models to replace the original model within the region of interest. These surrogate models can be constructed efficiently via reduced order modeling and subspace analysis. Once constructed, these surrogate models can be used to perform computationally expensive mathematical analyses. This work proposes a surrogate based model calibration algorithm. The proposed algorithm is used to calibrate various neutronics and thermal-hydraulics parameters. The virtual environment for reactor applications-core simulator (VERA-CS) is used to simulate a three-dimensional core depletion problem. The proposed algorithm is then used to construct a reduced order model (a surrogate) which is then used in a Bayesian approach to calibrate the neutronics and thermal-hydraulics parameters. The algorithm is tested and the benefits of data assimilation and calibration are highlighted in an uncertainty quantification study and requantification after the calibration process. Results showed that the proposed algorithm could help to reduce the uncertainty in key reactor attributes based on experimental and operational data.

• Asian Pacific Journal of Cancer Prevention
• /
• 제16권17호
• /
• pp.7785-7788
• /
• 2015

Utilization of high energy photons (>10MV) with an optimal weight using a mixed energy technique is a practical way to generate a homogenous dose distribution while maintaining adequate target coverage in intact breast radiotherapy. This study represents a model for estimation of this optimal weight for day to day clinical usage. For this purpose, treatment planning computed tomography scans of thirty-three consecutive early stage breast cancer patients following breast conservation surgery were analyzed. After delineation of the breast clinical target volume (CTV) and placing opposed wedge paired isocenteric tangential portals, dosimeteric calculations were conducted and dose volume histograms (DVHs) were generated, first with pure 6MV photons and then these calculations were repeated ten times with incorporating 18MV photons (ten percent increase in weight per step) in each individual patient. For each calculation two indexes including maximum dose in the breast CTV ($D_{max}$) and the volume of CTV which covered with 95% Isodose line ($V_{CTV,95%IDL}$) were measured according to the DVH data and then normalized values were plotted in a graph. The optimal weight of 18MV photons was defined as the intersection point of $D_{max}$ and $V_{CTV,95%IDL}$ graphs. For creating a model to predict this optimal weight multiple linear regression analysis was used based on some of the breast and tangential field parameters. The best fitting model for prediction of 18MV photons optimal weight in breast radiotherapy using mixed energy technique, incorporated chest wall separation plus central lung distance (Adjusted R2=0.776). In conclusion, this study represents a model for the estimation of optimal beam weighting in breast radiotherapy using mixed photon energy technique for routine day to day clinical usage.

• 한국전산구조공학회논문집
• /
• 제25권4호
• /
• pp.355-362
• /
• 2012

건전성 예측은 구조물의 고장이 발생될 때까지 남은 시간인 잔존유효수명을 예측하는 것으로, 이는 안전 및 정비 계획과 직접적으로 연관되기 때문에 매우 중요하다. 건전성 예측방법에는 물리모델 기반방법, 데이터 기반방법과 두 방법의 장점을 통합하는 방법이 있으며, 본 연구에서는 잔존수명 예측의 정확도가 모델변수 추정과 직접적으로 관련되는 물리모델 기반 건전성 예측에 초점을 맞춘다. 물리모델기반 건전성 예측에서는 모델변수 추정을 통해 시스템 상태의 장기 예측이 가능하지만, 대부분의 실제 구조물들의 상태모델은 여러 개의 모델변수를 포함함은 물론이고, 그 변수들이 서로 상관되어 있기 때문에 모델변수를 추정하는 일은 간단한 문제가 아니다. 본 연구에서는 물리모델 기반 건전성 예측을 위한 세 가지 변수 추정방법들의 차이를 논한다. 이 세 가지 방법들은 파티클 필터, 전반적인 베이지안 접근법, 그리고 순차적인 베이지안 접근법으로 모두 베이지안 추론이라는 하나의 이론적 바탕에 기반하지만, 샘플링 방법이나 갱신 절차 등에서 차이가 있다. 균열성장을 표현하는 Paris 모델의 변수 추정을 통해 세 가지 방법의 차이점이 논해지고, 건전성 예측 메트릭을 이용하여 정량적 차이를 표현한다. 파티클 필터방법이 건전성 예측 메트릭 측면에서 가장 높은 성능을 나타내었지만, 전반적인 베이지안 방법은 파티클 필터방법과 근소한 차이를 보이면서도 데이터가 집단으로 존재할 때에는 가장 효율적인 방법으로 나타났다.

• 한국의학물리학회지:의학물리
• /
• 제31권3호
• /
• pp.54-62
• /
• 2020

Dose calculation algorithms play an important role in radiation therapy and are even the basis for optimizing treatment plans, an important feature in the development of complex treatment technologies such as intensity-modulated radiation therapy. We reviewed the past and current status of dose calculation algorithms used in the treatment planning system for radiation therapy. The radiation-calculating dose calculation algorithm can be broadly classified into three main groups based on the mechanisms used: (1) factor-based, (2) model-based, and (3) principle-based. Factor-based algorithms are a type of empirical dose calculation that interpolates or extrapolates the dose in some basic measurements. Model-based algorithms, represented by the pencil beam convolution, analytical anisotropic, and collapse cone convolution algorithms, use a simplified physical process by using a convolution equation that convolutes the primary photon energy fluence with a kernel. Model-based algorithms allowing side scattering when beams are transmitted to the heterogeneous media provide more precise dose calculation results than correction-based algorithms. Principle-based algorithms, represented by Monte Carlo dose calculations, simulate all real physical processes involving beam particles during transportation; therefore, dose calculations are accurate but time consuming. For approximately 70 years, through the development of dose calculation algorithms and computing technology, the accuracy of dose calculation seems close to our clinical needs. Next-generation dose calculation algorithms are expected to include biologically equivalent doses or biologically effective doses, and doctors expect to be able to use them to improve the quality of treatment in the near future.