• 한국정보통신학회논문지
• /
• 제24권5호
• /
• pp.591-599
• /
• 2020

딥러닝 기술은 실세계의 객체를 분류하거나 인식하기 위해서 사용된다. 이를 위해서 준비된 많은 데이터를 고성능 컴퓨터에서 학습한 후에, 그 학습모델을 인식기에 탑재하여 각종 객체들을 인식한다. 이러한 인식기는 다양한 환경에서 사용되면서 인식하지 못하는 객체들이나 인식률이 낮은 객체들이 발생할 수 있다. 이런 문제를 해결하기 위해서 실세계 객체들을 주기적으로 학습하여 인식률을 높인다. 하지만, 즉각적인 인식률 향상이 어려울 뿐만 아니라, 임베디드 디바이스 등에 탑재되어 있는 인식기에서 학습하는 것이 쉽지 않다. 따라서, 본 논문에서는 임베디드 디바이스에 적용 가능한 부분 학습 기반의 실시간 손글씨 인식기를 제안한다. 제안된 인식기는 사용자 요청 시마다 임베디드 디바이스에서 부분 학습을 할 수 있는 환경을 제공하고, 실시간으로 인식기의 학습모델이 갱신된다. 이로 인해서 인식기의 지능이 지속적으로 향상됨으로 최초에 인식하지 못했던 손글씨에 대해 인식이 가능해진다. 이렇게 제안된 인식기는 RK3399 임베디드 디바이스에서 22개의 숫자와 글자에 대해서 학습과 추론이 가능하다는 것을 실험을 통하여 사람 손으로 쓴 은행 계좌명과 계좌번호를 인식할 수 있는 개인화된 지능을 가진 스마트 기기에 활용 가능할 것으로 기대된다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.1727-1731
• /
• 2003

In this paper we proposed backward fuzzy Petri-net to diagnoses faults in power systems by using the time sequence information of protection system. As the complexity of power systems increases, especially in the case of multiple faults or incorrect operation of protective devices, fault diagnosis requires new and systematic methods to the reasoning process, which improves both its accuracy and its efficiency. The fuzzy Petri-net models of protection system are composed of the operating process of protective devices and the fault diagnosis process. Fault diagnosis model, which makes use of the nature of fuzzy Petri-net, is developed to overcome the drawbacks of methods that depend on operator knowledge. The proposed method can reduce processing time and increase accuracy when compared with the traditional methods. And also this method covers online processing of real-time data from SCADA (Supervisory Control and Data Acquisition)

• Korean Journal of Hydrosciences
• /
• 제8권
• /
• pp.69-83
• /
• 1997

The Parameters of the storage function model (SFM) are taken as constants, while they have different values during every rainfall period and the duration of the runoff. Therefore, the results of the SFM generally show remarkably large errors. In this study, the modified storage function model (MSFM), in which the time variant parameters are introduced, is proposed to improve the SFM which is a conceptual rainfall-runoff model. The fuzzy reasoning method is applied as a real-time control one of the time variant parameters of the proposed model. The applicability of the MSFM was examined in the Bochung river, at a tributary of the Geum River, Korea. The pattern of the predicted runoff hydrograph and the peak discharge by the MSFM with fuzzy control are very similar to the measured values, compared with the results produced by the SFM.

• 유체기계공업학회:학술대회논문집
• /
• 유체기계공업학회 2003년도 유체기계 연구개발 발표회 논문집
• /
• pp.161-166
• /
• 2003

It is the common features of the integral reactors that the main components of the RCS are installed within the reactor vessel, and so there are no any flow pipes connecting the coolant pumps or steam generators. Due to no any flow pipes, it is impossible to measure the differential pressure at the RCS of the integral reactors, and it also makes impossible measure the flow-rate of the reactor coolant. As a alternative method, the method by the measurement of coolant pump power has been introduced in this study. Up to now, we did not found out a precedent which the coolant pump power is used for the real-time flow-rate calculation at normal operation of the commercial nuclear power plants. The objective of the study is to embody the real-time flow-rate calculation method by the measurement of coolant pump power in an integral reactor. As a result of the study, we could theoretically reason that the capacity-head curve and capacity-shaft power curve around the rated capacity with the high specific-speeded axial flow pumps have each diagonally steep incline but show the similar shape. Also, we could confirm the above theoretical reasoning from the measured result of the pump motor inputs, So, it has been concluded that it is possible to calculate the real-time flow-rate by the measurement of pump motor inputs. In addition, the compensation for a above new method can be made by HBM being now used in the commercial nuclear power plants.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제27권2호
• /
• pp.120-133
• /
• 2000

산업현장에서 복합 실시간 계통(HRTS: Hybrid Real-Time Systems) 개발을 위한 정형기법 사용의 주된 장벽은 인지적 어려움이며 이는 또 다른 위험을 초래할 수 있다. 이러한 문제를 극복하기 위해 HRTS 요구분석과 안전성 분석 시 사용자의 인지적 부담을 줄여줄 수 있는 정성적 요구분석 체계를 제안한다. 이 체계는 요구사항 명세를 위한 정성적 정형기법(QFM: Qualitative Formal Method)과 인과정보에 의한 요구사항 안전성 분석기법(CRSA: Causal Requirements Safety Analysis)으로 구성되어 있다. QFM에서는 인공지능 분야에서 연구된 정성추론 이론을 정형명세에 도입하여 요구사항 설계자와 분석자의 인지적 부담을 줄일 수 있도록 하였다. CRSA는 QFM에서 도출한 HRTS 동작의 인과 정보에 따라 체계적으로 위험 원인을 추적할 수 있도록 하여, 기존 결함 트리 분석(FTA: Fault Tree Analysis) 기법의 단점인 분석자의 주관에 의존하는 문제를 해결한다. 월성 원자력 발전소 자동정지계통(Shutdown System 2) 소프트웨어 요구사항 명세와 안전성 분석에 QFM과 CRSA를 적용하여 그 실효성을 입증하고자 하였다.

• Nuclear Engineering and Technology
• /
• 제55권12호
• /
• pp.4408-4425
• /
• 2023

Alarm flood due to abnormality propagation is the most difficult alarm overloading problem in nuclear power plants (NPPs). Root-cause analysis is suggested to help operators in understand emergency events and plant status. Multilevel Flow Modeling (MFM) has been extensively applied in alarm management by virtue of the capability of explaining causal dependencies among alarms. However, there has never been a technique that can identify the actual root cause for complex alarm situations. This paper presents an automated root-cause analysis system based on MFM. The causal reasoning algorithm is first applied to identify several possible root causes that can lead to massive alarms. A novel root-cause ranking algorithm can subsequently be used to isolate the most likely faults from the other root-cause candidates. The proposed method is validated on a pressurized water reactor (PWR) simulator at HAMMLAB. The results show that the actual root cause is accurately identified for every tested operating scenario. The automation of root-cause identification and ranking affords the opportunity of real-time alarm analysis. It is believed that the study can further improve the situation awareness of operators in the alarm flooding situation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4345-4363
• /
• 2021

Deep Learning as a Service (DLaaS), utilizing the cloud-based deep neural network models to provide customer prediction services, has been widely deployed on mobile cloud computing (MCC). Such services raise privacy concerns since customers need to send private data to untrusted service providers. In this paper, we devote ourselves to building an efficient protocol to classify users' images using the convolutional neural network (CNN) model trained and held by the server, while keeping both parties' data secure. Most previous solutions commonly employ homomorphic encryption schemes based on Ring Learning with Errors (RLWE) hardness or two-party secure computation protocols to achieve it. However, they have limitations on large communication overheads and costs in MCC. To address this issue, we present LeHE4SCNN, a scalable privacy-preserving and communication-efficient framework for CNN-based DLaaS. Firstly, we design a novel low-expansion rate homomorphic encryption scheme with packing and unpacking methods (LeHE). It supports fast homomorphic operations such as vector-matrix multiplication and addition. Then we propose a secure prediction framework for CNN. It employs the LeHE scheme to compute linear layers while exploiting the data shuffling technique to perform non-linear operations. Finally, we implement and evaluate LeHE4SCNN with various CNN models on a real-world dataset. Experimental results demonstrate the effectiveness and superiority of the LeHE4SCNN framework in terms of response time, usage cost, and communication overhead compared to the state-of-the-art methods in the mobile cloud computing environment.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.1039-1043
• /
• 2005

Despite the many significant advances made in robot architecture, the basic approaches are deliberative and reactive methods. They are quite different in recognizing outer environment and inner operating mechanism. For this reason, they have almost opposite characteristics. Later, researchers integrate these two approaches into hybrid architecture. In such architecture, Reactive module also called low-level motion control module have advantage in real-time reacting and sensing outer environment; Deliberative module also called high-level task planning module is good at planning task using world knowledge, reasoning and intelligent computing. This paper presents a framework of the integrated planning and control for mobile robot navigation. Unlike the existing hybrid architecture, it learns topological map from the world map by using MST (Minimum Spanning Tree)-based SOFM (Self-Organizing Feature Map) algorithm. High-level planning module plans simple tasks to low-level control module and low-level control module feedbacks the environment information to high-level planning module. This method allows for a tight integration between high-level and low-level modules, which provide real-time performance and strong adaptability and reactivity to outer environment and its unforeseen changes. This proposed framework is verified by simulation.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2000년도 하계학술대회 논문집 A
• /
• pp.122-124
• /
• 2000

As the complexity of power systems increases, especially in the case of multiple faults or incorrect operation of protective devices, fault diagnosis requires new and systematic methods to the reasoning process, which improves both its accuracy and its efficiency. Therefore this paper proposes a method of the modeling of protection systems and fault diagnosis in power systems using Fuzzy Petri Nets (FPN). The proposed method can reduce processing time and increase accuracy when compared with the traditional methods. And also this method can cover online processing of real-time data from SCADA (Supervisory Control and Data Acquisition).

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2000년도 제15차 학술회의논문집
• /
• pp.15-15
• /
• 2000

In the past couple of years, there has been increasing interest in the fusion of neural networks and fuzzy logic. Most of the existing fused models have been proposed to implement different types of fuzzy reasoning mechanisms and inevitably they suffer from the dimensionality problem when dealing with complex real-world problem. To overcome the problem, we propose the self-organizing networks with activation nodes based on fuzzy inference and polynomial function. The proposed model consists of two parts, one is fuzzy nodes which each node is operated as a small fuzzy system with fuzzy implication rules, and its fuzzy system operates with Gaussian or triangular MF in Premise part and constant or regression polynomials in consequence part. the other is polynomial nodes which several types of high-order polynomials such as linear, quadratic, and cubic form are used and are connected as various kinds of multi-variable inputs. To demonstrate the effectiveness of the proposed method, time series data for gas furnace process has been applied.