• Nuclear Engineering and Technology
• /
• 제54권5호
• /
• pp.1644-1651
• /
• 2022

This paper focuses on the power-level control of nuclear power plants (NPPs) in the presence of lumped disturbances. An adaptive second-order nonsingular terminal sliding mode control (ASONTSMC) scheme is proposed by resorting to the second-order nonsingular terminal sliding mode. The pre-existing mathematical model of the nuclear reactor system is firstly described based on point-reactor kinetics equations with six delayed neutron groups. Then, a second-order sliding mode control approach is proposed by integrating a proportional-derivative sliding mode (PDSM) manifold with a nonsingular terminal sliding mode (NTSM) manifold. An adaptive mechanism is designed to estimate the unknown upper bound of a lumped uncertain term that is composed of lumped disturbances and system states real-timely. The estimated values are then added to the controller, resulting in the control system capable of compensating the adverse effects of the lumped disturbances efficiently. Since the sign function is contained in the first time derivative of the real control law, the continuous input signal is obtained after integration so that the chattering effects of the conventional sliding mode control are suppressed. The robust stability of the overall control system is demonstrated through Lyapunov stability theory. Finally, the proposed control scheme is validated through simulations and comparisons with a proportional-integral-derivative (PID) controller, a super twisting sliding mode controller (STSMC), and a disturbance observer-based adaptive sliding mode controller (DO-ASMC).

• 한국군사과학기술학회지
• /
• 제27권4호
• /
• pp.437-445
• /
• 2024

In future complex and diverse battlefield situations, the existing command system faces the challenge of delayed human judgement of strategy and low objectivity. This paper proposes an artificial intelligence model that takes situation information and course of action simulation results as input and automatically assigns scores to various evaluation elements and a comprehensive score. This tool is expected to assist the commander in making decisions, reduce the time required for making judgments, and promote impartial decision-making.

• 한국산학기술학회논문지
• /
• 제22권3호
• /
• pp.58-64
• /
• 2021

자동차 등록대수와 비례하여 증가하는 교통 혼잡은 도시의 사회경제 발전의 저해 요소로 작용하고 있다. 본 논문은 VDS(Vehicle Detection System)을 통한 데이터를 입력 변수로 사용한다. 본 연구의 목적은 교통 흐름을 단순히 2단계(원할, 정체)가 아닌 5단계(원할, 다소 지체, 지체, 다소 정체, 정체)로 더 정교하게 예측하고, 이 예측에서 가장 정확도가 높은 모델인 Catboost 모델과 다른 모델들을 비교하는 것이다. 이를 위해 본 논문에서는 머신러닝 알고리즘인 Catboost 모델을 통해 5가지 단계를 예측하고 정확도를 다른 머신러닝 알고리즘들과 비교, 분석한다. 또한, 하이퍼 파라미터(Hyper Parameter) 튜닝 및 원-핫 인코딩(One-Hot Encoding) 전처리를 거치지 않은 Catboost 모델과 랜덤 선택(RandomizedSearchcv)을 통해 튜닝 및 데이터 전처리를 거친 모델을 비교, 분석한다. 분석 결과 하이퍼 파라미터 튜닝을 하지 않은 초기 Catboost 모델이 정확도 93%를 보이며 가장 높은 정확도를 기록하였다. 따라서 본 연구는 두가지 의의를 가진다. 첫번째로, 초기 세팅된 파라미터들이 적용된 Catboost 모델이 다수의 범주형 변수를 포함하는 교통 흐름 예측에서 다른 머신러닝, 딥러닝 모델들보다 성능이 높다는 결론을 도출했다는 점에서 의의가 있다. 두번째로, 기존 2단계로 예측하던 교통 흐름을 5단계로 예측함으로써 더욱 정교한 교통 흐름 예측 모델을 제안한다는 점에서 의의를 가진다.

• 한국항공우주학회지
• /
• 제37권4호
• /
• pp.350-358
• /
• 2009

본 논문은 재진입 비행체의 TAEM 구간 유도와 제어에 관하여 기술 하였다. TAEM 구간은 공기의 밀도와 비행체의 속도의 범위가 큰 특징을 가지고 있으며, 이들 조건하에 TAEM 구간의 유도를 위한 궤적과 비행체의 상태값을 최적화하였다. 최적화된 상태값은 7가지의 상태이며, 상태값은 Down-range, Cross-range, 비행체의 고도, 속도, 경로각, 방위각, 그리고 비행 거리이다. 최적화 연산을 수행하기 위하여 DIDO 프로그램을 사용하였다. 재진입 비행체의 제어를 위하여 인공 신경망을 이용한 되먹임 선형화 제어법을 사용하였다. 비행체의 동역학 모델은 역전파 모델을 통하여 근사화 되고, 근사화된 동역학 모델과 지연된 제어 입력, 플랜트 출력으로부터 새로운 제어 입력을 생성하게 된다. 이를 이용하여 본 논문에서는 앞서 최적화된 7가지의 상태값을 추종하는 결과를 보였다.

• 한국의료질향상학회지
• /
• 제7권1호
• /
• pp.106-116
• /
• 2000

Background : Severance Hospital is an university hospital which has 1,580 beds. A LAN system was installed in the Medical Record Department in 1992 and discharge abstract data have been added to the discharge abstract database(DB) The previous work flow in the Medical Record Department had 5 levels: 1) chart collection from wards, 2) assembling, 3) abstracting data from medical record on worksheet by 2 RRAs, 4) checking deficiencies and coding diagnosis and procedures by 4 RRAs, 5) inputting the data into the discharge abstract data base by 1 RRA. The average processing time took 19.3 days from the patient discharge date. It had the production of monthly statistical report delayed. Besides, it caused the users in the hospital to complain. Methods : A CQI team was organized to find a way to shorten the processing time less than 10 days. The team identified the factors making the processing time long and integrated three levels from the 3rd level into one. Each of 7 RRAs performed the integrated level on her workstation instead of taking one of three separate levels. The comparison of processing time before and after the changes was made with 3'846 discharges of April, 1999 and 4,189 discharges of August, 1999. Results : The average processing time was shortened from 19.3 days to 8.7 days. Especially the integrated level took only 3.6 days, compared with 12.3 days before the change. The percentage of finishing up the whole processing within 10 days from discharge was increased up to 77.6%, which was 2.4% before the integration. The prevalence of error in data input was not increased in the new method. Conclusions : The integrated processing method has the following advantages: 1) the expedition of production of monthly statistical report, 2) the increase of utilizing rate of dischare abstract data by Billing Dept, Emergency Room, QI Dept., etc., 3) the improvement of intradepartmental work follow, 4) the enhancement of medical record quality by checking the deficiencies earlier than before.

• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2001년도 The Pacific Aisan Confrence On Intelligent Systems 2001
• /
• pp.59-64
• /
• 2001

The robot soccer simulation game is a dynamic multi-agent environment. In this paper we suggest a new reinforcement learning approach to each agent's dynamic positioning in such dynamic environment. Reinforcement learning is the machine learning in which an agent learns from indirect, delayed reward an optimal policy to chose sequences of actions that produce the greatest cumulative reward. Therefore the reinforcement learning is different from supervised learning in the sense that there is no presentation of input pairs as training examples. Furthermore, model-free reinforcement learning algorithms like Q-learning do not require defining or learning any models of the surrounding environment. Nevertheless it can learn the optimal policy if the agent can visit every state- action pair infinitely. However, the biggest problem of monolithic reinforcement learning is that its straightforward applications do not successfully scale up to more complex environments due to the intractable large space of states. In order to address this problem. we suggest Adaptive Mediation-based Modular Q-Learning (AMMQL)as an improvement of the existing Modular Q-Learning (MQL). While simple modular Q-learning combines the results from each learning module in a fixed way, AMMQL combines them in a more flexible way by assigning different weight to each module according to its contribution to rewards. Therefore in addition to resolving the problem of large state effectively, AMMQL can show higher adaptability to environmental changes than pure MQL. This paper introduces the concept of AMMQL and presents details of its application into dynamic positioning of robot soccer agents.

• 한국작물학회지
• /
• 제44권3호
• /
• pp.221-224
• /
• 1999

A field experiment was carried out to observe the weed control effects of vetch and to evaluate vetch characteristics on clay loam soil in no-till direct-sown rice-vetch cropping systems. The effects of weed control, forage productivity, and N content of vetch plants were investigated. With the progress of plant growth, density of Chinese milk vetch (Astragalus sinicus L.) gradually decreased, but densities of foxtail and other weeds decreased steeply due to the depression by the over-shaded vetch canopy in a no-till direct-sown rice-vetch cropping paddy field. The vetch density in tillage systems was lower than in notillage cropping systems. Lower vetch density occurred with an increase in foxtail density and other weeds. Weed control effect increased by the progress of vetch growth, which indicated that the vetch canopy over-shaded the weeds. Vetch straw was degraded rapidly submergence after with water at the time of wet sowing of rice. Early harvesting of vetch seed resulted in lower seed germination. To acquire enough seedlings without re-sowing, the harvesting of seed should be delayed at least 28 days after the flowering stage in order to ensure the vetch population is sustainable in a no-till direct-sown rice-vetch cropping system. In order to improve the survival of vetch plants, vetch seeds should germinate from the heading .stage to before the full-ripening stage of rice plants. To enhance the percent of over-wintering survival, vetch seeds should germinate no later than the end of October in southern Korea. The dry weight of vetch plants increased with the progress of vetch growth until the flowering stage but N content decreased for 30 days from before the flowering stage (2.9%) until the ripening stage (1.8%). We concluded that Chinese milk vetch could have an effect on weed control before the flowering stage, sustainability without re-sowing of seed annually, and effective green manure for rice pre-crop in no-till direct-sown rice-vetch relay cropping systems.

• 한국시뮬레이션학회:학술대회논문집
• /
• 한국시뮬레이션학회 2001년도 The Seoul International Simulation Conference
• /
• pp.321-324
• /
• 2001

The robot soccer simulation game is a dynamic multi-agent environment. In this paper we suggest a new reinforcement learning approach to each agent's dynamic positioning in such dynamic environment. Reinforcement Beaming is the machine learning in which an agent learns from indirect, delayed reward an optimal policy to choose sequences of actions that produce the greatest cumulative reward. Therefore the reinforcement loaming is different from supervised teaming in the sense that there is no presentation of input-output pairs as training examples. Furthermore, model-free reinforcement loaming algorithms like Q-learning do not require defining or loaming any models of the surrounding environment. Nevertheless it can learn the optimal policy if the agent can visit every state-action pair infinitely. However, the biggest problem of monolithic reinforcement learning is that its straightforward applications do not successfully scale up to more complex environments due to the intractable large space of states. In order to address this problem, we suggest Adaptive Mediation-based Modular Q-Learning(AMMQL) as an improvement of the existing Modular Q-Learning(MQL). While simple modular Q-learning combines the results from each learning module in a fixed way, AMMQL combines them in a more flexible way by assigning different weight to each module according to its contribution to rewards. Therefore in addition to resolving the problem of large state space effectively, AMMQL can show higher adaptability to environmental changes than pure MQL. This paper introduces the concept of AMMQL and presents details of its application into dynamic positioning of robot soccer agents.

• 한국항공우주학회지
• /
• 제39권7호
• /
• pp.618-626
• /
• 2011

무인항공기가 복잡한 지형이나 위험 지역에서 안전한 비행을 하기 위해서는 요구되는 경로를 정밀하게 추종할 수 있는 제어기법이 필요하다. 경로추종을 위해 사용되는 PID 제어기에서는 경로의 곡률을 앞먹임하여 추종성능을 개선할 수 있다. 한편 경로를 정밀 추종하기 위해서는 제어기의 이득을 크게 하는 것이 필요하나 비행체의 응답속도가 느린 경우 큰 경로 추종 이득을 사용할 경우 성능 저하 또는 불안정성이 발생할 수 있다. 여기서는 응답지연을 고려하여 앞먹임을 갖는 PID제어기의 설계방법을 고려하였다. 앞먹임에 필요한 경로정보를 간단히 얻기 위해 주어진 경로를 3차 스플라인 방법을 적용하여 3차 다항식으로 나타내었다. 설계한 제어기의 추종성능을 평가하기 위해 높은 고도에서 운용되는 느린 횡방향 동특성을 갖는 무인항공기에 대해 비선형 시뮬레이션을 수행하였다. 제어기에서는 횡방향 동특성을 1차 모델로 가정하여 반영하였다. 시뮬레이션을 통하여 동특성을 고려한 경우는 비행체가 주어진 경로를 매우 정밀하게 추종함을 확인하였다.

• 대기
• /
• 제23권3호
• /
• pp.275-282
• /
• 2013

In this study, the numerical model was developed to evaluate the observational environment of sunshine duration and, for evaluating the accuracy and utility of the model, it was verified against the observational data measured at Dae-gu Automated Synoptic Observing System (ASOS) located in an urban area. Three-dimensional topography and building configuration as the surface input data of the model were constructed using a Geographic Information System (GIS) data. First, the accuracy of the computing planetary positions suggested by Paul Schlyter was verified against the data provided by Korea Astronomy and Space Science Institute (KASI) and the results showed that the numerical model predicted the Sun's position (the solar azimuth and altitude angles) quite precisely. Then, this model was applied to reproduce the sunshine duration at the Dae-gu ASOS. The observed and calculated sunshine durations were similar to each other. However, the observed and calculated sunrise (sunset) times were delayed (curtailed), compared to those provided by KASI that considered just the ASOS's position information such as latitude, longitude, and elevation height but did not consider the building and topography information. Further investigation showed that this was caused by not only the topographic characteristic (higher in the east and lower in the west) but also the buildings located in the southeast near the sunrise and the southwest near the sunset. It was found that higher building resolution increased the accuracy of the model. It was concluded that, for the accurate evaluation of the sunshine duration, detailed building and topography information around the observing sites was required and the numerical model developed in this study was successful to predict and/or the sunshine duration of the ASOS located in an urban area.