• 한국전자통신학회논문지
• /
• 제18권6호
• /
• pp.1331-1336
• /
• 2023

머신러닝을 이용한 텍스트 기반 어류 질병 분류에서 머신러닝 모델의 입력 파라미터가 너무 많은 문제가 존재하지만, 성능의 문제로 임의로 입력 파라미터를 줄일 수 없다. 본 논문에서는 이 문제를 해결하고자 SHAP 분석 기법을 활용해 넙치 질병 분류에 특화된 입력 파라미터 최적화 방안을 제시한다. 제안한 방법은 SHAP 분석 기법을 적용하여 넙치 질병 문진표에서 추출한 질병 정보의 데이터 전처리와 AutoML을 활용한 머신러닝 모델 평가 과정을 포함한다. 이를 통해 AutoML의 입력 파라미터의 성능을 평가하고, 최적의 입력 파라미터 조합을 도출한다. 본 연구에서 제안 방법은 필요한 입력 파라미터 수를 감소시키면서도 기존의 성능을 유지할 수 있을 것으로 기대되며, 이는 텍스트 기반 넙치 질병 분류의 효율성 및 실용성을 높이는 데 기여할 것이다.

• 한국수산과학회지
• /
• 제57권3호
• /
• pp.227-238
• /
• 2024

This study aimed to evaluate the effects of polyethylene microplastic (PE-MPs) via measuring the growth performance, hematological parameters, and antioxidant responses in Korean rockfish Sebastes schlegelii exposed to waterborne polyethylene microplastic with dimensions of 22-71 ㎛. S. schlegelii (mean weight, 34.55±5.82 g; mean length, 12.59±0.79 cm) were exposed to PE-MPs at concentrations of 0, 400, 800, 1,600 and 3,200 ㎍/L for 10 and 20 days. PE-MPs significantly affected growth performance, hematological parameters, plasma components, and antioxidant responses in a concentration-dependent manner. At a concentration ≥1,600 ㎍/L, PE-MPs significantly decreased body weight gain and specific growth rate, and significantly increased the hepatosomatic index. Hematological parameters showed a significant decrease in total red blood cell count and hemoglobin levels. Plasma components showed a significant increase in glucose, aspartate aminotransferase, and alanine transaminase levels, whereas total protein, calcium, and magnesium levels significantly decreased. Exposure to ≥1,600 ㎍/L PE-MPs also induced reactive oxygen species generation in the gill and liver, significantly increasing superoxide dismutase and catalase activity. These findings suggest that exposure to ≥1,600 ㎍/L PE-MPs could significantly change growth performance, hematological parameters, plasma components, and antioxidant responses, resulting in physiological toxicity.

• Smart Structures and Systems
• /
• 제13권3호
• /
• pp.473-500
• /
• 2014

Tuned mass dampers (TMDs) have been installed in many high-rise buildings, to improve their resiliency under dynamic loads. However, high-rise buildings may experience natural frequency changes under ambient temperature fluctuations, extreme wind loads and relative humidity variations. This makes the design of a TMD challenging and may lead to a detuned scenario, which can reduce significantly the performance. To alleviate this problem, the current paper presents a proposed approach for the design of a robust and efficient TMD. The approach accounts for the uncertain natural frequency, the optimization objective and the input excitation. The study shows that robust design parameters can be different from the optimal parameters. Nevertheless, predetermined optimal parameters are useful to attain design robustness. A case study of a high-rise building is executed. The TMD designed with the proposed approach showed its robustness and effectiveness in reducing the responses of high-rise buildings under multidirectional wind. The case study represents an engineered design that is instructive. The results show that shear buildings may be controlled with less effort than cantilever buildings. Structural control performance in high-rise buildings may depend on the shape of the building, hence the flow patterns, as well as the wind direction angle. To further increase the performance of the robust TMD in one lateral direction, active control using LQG and fuzzy logic controllers was carried out. The performance of the controllers is remarkable in enhancing the response reduction. In addition, the fuzzy logic controller may be more robust than the LQG controller.

• 대한전기학회논문지:전력기술부문A
• /
• 제48권6호
• /
• pp.789-799
• /
• 1999

In this paper, the optimization of fuzzy inference systems is proposed for fuzzy model of nonlinear systems. A fuzzy model needs to be identified and optimized by means of the definite and systematic methods, because a fuzzy model is primarily acquired by expert's experience. The proposed rule-based fuzzy model implements system structure and parameter identification using the HCM(Hard C-mean) clustering method, genetic algorithms and fuzzy inference method. Two types of inference methods of a fuzzy model are the simplified inference and linear inference. in this paper, nonlinear systems are expressed using the identification of structure such as input variables and the division of fuzzy input subspaces, and the identification of parameters of a fuzzy model. To identify premise parameters of fuzzy model, the genetic algorithms is used and the standard least square method with the gaussian elimination method is utilized for the identification of optimum consequence parameters of fuzzy model. Also, the performance index with weighting factor is proposed to achieve a balance between the performance results of fuzzy model produced for the training and testing data set, and it leads to enhance approximation and predictive performance of fuzzy system. Time series data for gas furnace and sewage treatment process are used to evaluate the performance of the proposed model.

• 한국생산제조학회지
• /
• 제24권2호
• /
• pp.217-223
• /
• 2015

In this paper, the performance of a system was analyzed according to the design parameters of a LED secondary lens that can be applied at a very close range, e.g., for direct lighting or display systems. We designed the secondary lens of the very-close-range LED using an aspheric equation and analyzed its performance-particularly the angle of the beam spread, central luminous intensity, and light uniformity-with respect to the thickness of lens, radius, conic constant, and asphericity (4th). Our analysis shows that four parameters affect the performance. The simulation results indicate an optimal thickness of 1 mm and show that a larger radius yields higher performance. The optimal range for the conic constant was determined as -1.21 to -1.25, the optimal range for the asphericity was determined as 0.0047xx to 0.0049xx (4th).

• 한국공작기계학회논문집
• /
• 제16권4호
• /
• pp.71-78
• /
• 2007

Input shaping is known to be a very effective tool for suppressing residual vibration without introducing any complicated sensors and feedback control. Real-time input shaping schemes necessitate a process such that the input command is discretized to deal with non-prescribed, real-time input. Thus parameters associated with input command discretization, such as time spacing and duration time, are unknowns which affect the performance of input shaping schemes, especially for small and fast XY stages. This paper investigates the effects of input command discretization parameters, such as time spacing and duration time, on the dynamic performance of XY stages subjected to real-time input shaping. An experimental system is developed which is equipped with an XY stage driven by servo-motors and real-time user command. Experiments are performed to investigate the dynamic performance of XY stage by changing these parameters and to yield a strategy to gain better performance.

• 상하수도학회지
• /
• 제23권4호
• /
• pp.417-430
• /
• 2009

A BNR comprehensive performance evaluation (BNR CPE) system was established employing system-oriented evaluation methodology for biological nutrient removal (BNR) processes based on the CPE techniques developed by U.S. EPA for evaluation of conventional biological processes. The BNR CPE system applied to five domestic BNR plants adopting $A^2/O$ process confirmed that all target plants except the smallest one had not any serious defective performance and process stability was enhanced with increasing plant size. The system also clearly verified relatively poor performances in anoxic reactors without exception mainly due to influent carbon limit rather than functional defect. Consistent good performances were confirmed even during both winter season and wet weather generally known to be difficult to achieve satisfactory removals. Presentation of evaluation results by modified radar chart system simplified and clarified the evaluation and analysis procedures. The BNR CPE system could not only discover readily the causes of present and prospective poor performances but also facilitate the suggestion of their optimization options. Mutual effect and cause-and-effect among operation parameters and unit processes were also found easily using the evaluation system. The system justified that the adverse effect of defective operating parameters could be compensated by other favorable parameters, especially in anaerobic and anoxic reactors as well as during the winter season.

• Nuclear Engineering and Technology
• /
• 제36권4호
• /
• pp.338-345
• /
• 2004

This study examines the sensitivity of several fabrication parameters for dry process fuel, using a random sampling technique. The in-pile performance of dry process fuel with irradiation was calculated by a modified ELESTRES code, which is the CANDU fuel performance code system. The performance of the fuel rod was then analyzed using a Monte Carlo simulation to obtain the uncertainty of the major outputs, such as the fuel centerline temperature, the fission gas pressure, and the plastic strain. It was proved by statistical analysis that for both the dry process fuel and the $UO_2$ fuel, pellet density is one of the most sensitive parameters, but as for the fission gas pressure, the density of the $UO_2$ fuel exhibits insensitive behavior compared to that of the dry process fuel. The grain size of the dry process fuel is insensitive to the fission gas pressure, while the grain size of the $UO_2$ fuel is correlative to the fission gas pressure. From the calculation with a typical CANDU reactor power envelop, the centerline temperature, fission gas pressure, and plastic strain of the dry process fuel are higher than those of the $UO_2$ fuel.

• 한국정보전자통신기술학회논문지
• /
• 제11권6호
• /
• pp.658-663
• /
• 2018

본 논문에서는 콘텐츠 큐레이션 플랫폼의 성능평가 알고리즘을 제안하였다. 성능평가는 초기조건과 종료조건을 설정하고, 4개의 파라미터를 측정하여 수행하였다. 성능평가에 사용되는 파라미터는 서비스 응답속도, 응답결과 정확도, 분석결과 정밀도, 분석결과 재현율이다. 제안한 성능평가 알고리즘의 유효성을 확인하기 위해 118,738건의 데이터베이스를 직접 구축한 후, 인터넷에서 콘텐츠를 오픈하여 이를 클릭한 사용자 행위를 기반으로 실험을 수행하였다. 실험에 사용된 페이지 뷰는 6,975,365건이었다, 실험결과, 4개의 파라미터를 이용하면 콘텐츠 큐레이션 플랫폼의 종합평가와 단위 평가가 객관적으로 수행될 수 있음을 확인하였다.

• Journal of the Korean Data and Information Science Society
• /
• 제25권1호
• /
• pp.255-261
• /
• 2014

Many studies have estimated a mixture of binomial distributions. This paper considers an extension, a mixture of shifted binomial distributions, and the estimation of the distribution. The range of each component binomial distribution is rst evaluated and then for each possible value of shifted parameters, the EM algorithm is employed to estimate those parameters. From a set of possible value of shifted parameters and corresponding estimated parameters of the distribution, the likelihood of given data is determined. The simulation results verify the performance of the proposed method.