초록 ▼

가. 초기사건, 사고 시나리오 모사 및 시뮬레이션, 그리고 사고해석 DB 구축
- 대상의 안전필수계통 분석, 평가 및 선정
- 선제진단 필수정보 선정, 선제진단 방향 및 범위 설정
- 다양한 사고 초기사건, 시나리오 평가 및 선정(안전계통 작동 여부 포함)
- 사고해석 코드 초기사건, 사고 시나리오 모사
- 안전계통 작동성을 반영한 시뮬레이션

나. 데이터-기반 스마트센싱 모델 개발
- 타분야 스마트센싱 적용 조사 분석
- 안전필수계통 상태 선제진단용 필수정보 분석 및 선정
-

가. 초기사건, 사고 시나리오 모사 및 시뮬레이션, 그리고 사고해석 DB 구축
- 대상의 안전필수계통 분석, 평가 및 선정
- 선제진단 필수정보 선정, 선제진단 방향 및 범위 설정
- 다양한 사고 초기사건, 시나리오 평가 및 선정(안전계통 작동 여부 포함)
- 사고해석 코드 초기사건, 사고 시나리오 모사
- 안전계통 작동성을 반영한 시뮬레이션

나. 데이터-기반 스마트센싱 모델 개발
- 타분야 스마트센싱 적용 조사 분석
- 안전필수계통 상태 선제진단용 필수정보 분석 및 선정
- 선제진단 방향/범위 설정
- 사고해석 데이터-기반 스마트센싱 방법론 분석
- 스마트센싱 알고리즘 검토, 비교분석 및 평가

다. 광계측 측정 시스템 구축
- 배관계통에 발생하는 다양한 누설 생성 기구의 분석과 실험대상 선정
- 전단간섭계와 적외선 열화상 시스템을 이용한 누설 진단 기초 실험
- 다양한 고장 원인에 의한 밸브의 누설 탐지용 배관계통 설계, 제작
- 밸브 및 배관 누설 진단/측정 DB 구축

라. 실시간 안전필수계측기 고장진단 방법론 및 모델선정
- 원전 사고 시 정확한 사고파악과 적절한 대처를 위해 필요한 계측기 고장진단 방법론 및 모델 선정


마. 실시간 안전필수계측기 고장진단 알고리즘 개발 및 구현
- 선정된 진단 모델과 방법론을 기반으로 상세 알고리즘을 개발하고 PC 환경에서 구현하여 고장진단 성능을 평가함.
- 데이터-기반 고장진단 모델에 대한 알고리즘 PC 구현
- 정상운전 데이터와 설계기준 사고 데이터를 생산하여 정상운전 시와 사고시 데이터-기반 모델의 고장진단 성능 평가
- 물리적 기반 고장진단의 타당성 확인을 위해 비상노심냉각계통의 단순화 모델을 개발한 후 계측기 오동작에 대한 진단방안 시뮬레이션

바. 모니터링 대상 안전필수기기 선정 및 기계적 특성 분석
- 다양한 안전필수기기 중에서 모니터링이 필요한 안전필수기기를 선정하고 선정된 기기의 기계적 특성 분석함.
- 선정된 기기의 고장 모드를 분석하고, 분석 결과를 통해 선정된 기기의 고장 원인 및 종류를 분석함.
- 분석된 고장모드와 관련되는 계측기를 조사하여 일어날 수 있는 이상신호를 조사 및 분석함.


사. 선정된 안전필수기기의 건전성 평가 방법론 개발
- 선정된 기기의 이상상태를 모델을 선정함
- 선정된 고압안전주입펌프의 건전성상태 평가 방법론을 개발하기 위해서 다양한 방법론을 검토함.
- MFM 방법을 선정하고, 선정된 고압안전주입펌프의 모델링을 수행함.
- 고압안전주입펌프의 이상상태 모델을 기준으로 비정상 상태를 다양한 case로 가정하고, MFM Workbench 프로그램을 이용하여 시뮬레이션을 수행함.
- 시뮬레이션 결과를 통해 MFM 성능을 평가함.

아. 인력양성
- 전문인력양성 체계구축(전문가 확보, 기관별 전문교육 등)
- 협력 구축을 바탕으로 현장파견 실무 강의, 논문지도 등 맞춤형 전문 인력 양성
- 원자력산업 및 I&C 전문 인력양성을 위해 국제 워크숍, 실무자 강의, 해외 전문가 초청 강의 등의 다양한 프로그램을 수행함.

( 출처 : 보고서 요약서 5p )

Abstract ▼

Ⅳ. The Results of the R&D
< CHOSUN UNIVERSITY >
1. Establishment of the NPP accident analysis DB
To establish the accident analysis DB for early diagnosis of safety-critical systems, a lot of scenarios including various break sizes, positions and operating conditions(safety system actuation

Ⅳ. The Results of the R&D
< CHOSUN UNIVERSITY >
1. Establishment of the NPP accident analysis DB
To establish the accident analysis DB for early diagnosis of safety-critical systems, a lot of scenarios including various break sizes, positions and operating conditions(safety system actuation and their delayed operation) starting from LOCA were selected, analyzed and evaluated.

And the simulation for various break positions and many break sizes was performed for establishment of the accident analysis DB for early diagnosis of SCSs.
These data have been used to develop and verify the smart-sensing algorithms.

2. Development of the data-based smart-sensing models
To develop the data-based smart-sensing algorithm for early diagnosis of SCSs,in this project several intelligence techniques were reviewed. From the review,FNN(Fuzzy Neural Network), GMDH(Group Method of Data Handing), and SVR(Support Vector Regression) methodologies were proposed and the developed algorithms were optimized using optimization techniques including the genetic algorithm. The developed algorithms accurately predicted essential information for managing the accident effectively such as reactor vessel water level, hydrogen concentration in containment, and coolant leakage rate from break. These algorithms were verified using the accident analysis DB.

3. Construction of optical measurement systems
The cause and shape of wall thinning defect which is usually generated in the pipelines in NPPs and plant industry was analyzed and various leakage mechanism being generated in the pipeline system was analyzed and the optical measurement experiment was selected for the test object. A shearography interferometer with single-beam laser to measure the out-of-plane deformation was constructed in order to measure the leakage of NPPs pipelines using an optical nondestructive testing technique. For the precise measurement, a software to process images acquired by CCD camera was developed. Fundamental leakage measurement experiments to assume that a defect is included in the welded line of pipeline made of STS304 were conducted for the leakage diagnosis using the shearography and infrared thermography systems.

4. Nuclear I&C manpower training through university-research institute collaboration
Dispatch of graduate students to institutes, practical lessons of institute experts, financial support to graduate students for outside special lessons, and thesis joint advice were performed. And the nuclear engineering department of Chosun university opened collaborative lectures with KAERI experts for two semesters. The courses of collaborative lecture are as follows:
- Failure analysis and diagnostics (the second semester, 2013)
- Evaluation of equipment integrity (the first semester, 2014)

< KAERI >
1. Selection of Failure Diagnosis Methods and Models for Real-time Safety-critical Instruments

Through the research on the selection of instrument failure diagnostic methods and models in order to grasp the accurate behavior of accidents and to take an appropriate countermeasure, we have obtained the following results.
- Establishment of the development strategies through the reviews on the related domestic/international case studies, related codes/standards/regulating requirements for instrument failure diagnosis and countermeasure methods in NPPs.
- Selection of data-driven models applicable to accidents through the analysis on the strengths and weaknesses of the models, by means of review and analysis for the traditional failure diagnostic models
- Selection of the critical instruments required to failure diagnosis through the analysis of SAMG, Fukushima Daiichi nuclear accident in Japan, and the parameters on the safety parameter monitoring systems in domestic NPPs.
- We have preliminarily evaluated the applicability of models in the case of accidents through the review on the interrelationships among parameters of selected instruments by using a data-driven model. The result has been drawn that in the case of an accident, it should be required to obtain various training data sets adequate to a specific accident characteristics, after technically grouping the accident categories.
- Due to the limitations on the data-driven models, it is recommended to use combined methodologies that include data-driven models, physical models, traditional redundancy/diversity evaluation methods, and analytic methods for various signal trends.

2. Development and Implementation of Failure Diagnostic Algorithms for Real-time Safety-critical Instruments
We have developed specific algorithms based on the selected diagnostic models and methods, and then implemented on a PC-based environment, and evaluated the performance of the failure diagnosis. We have obtained the following results.
- Data acquisition and adjustment, grouping of variables, model design,optimization, and analysis of model performance and uncertainties, based on AAKR model. Implementation of failure detection algorithms by using a PC-based MATLAB.
- We have generated both normal operation data and design-basis accident data by using the KAERI CNS. As a result of evaluating the performance of data-driven model based failure diagnosis for both normal operation and accident condition, the following results have been drawn.

• In the case of using a data-driven model for normal operation, the performance of failure diagnosis shows good results, regardless of failure types.
• In the case of using the trained model of normal operational data set for design-basis accidents, the performance of failure diagnosis at the early stage of the accident shows reasonably good results, but after the mid-time stage of the accident, the performance is abruptly decreased.
• In the case of using a trained model of accident data set for design-basis accidents, the performance of instrument failure diagnosis shows good results, but for scenarios with some variations of the same accident type, the performance is decreased.
• Thus, in order to apply data-driven models, it should be necessary to train the models by grouping the accident classifications and scenarios appropriately.
• In order to verify the appropriateness of physical models, we have developed a simplified model for ECCS.

- By using the developed simplified model of ECCS, we have simulated the model to evaluate failure diagnostic capabilities on the instrument malfunctions. Thus, we have obtained the following results.
• The physical diagnostic model is a practical means because this model include the knowledge in terms of plant systems and the mirroring concept of those plant systems.
• Because the model shows clear interrelationships among instruments, the various states of equipment/process, this model can easily detect the instrument failures in a system.
• The applicability of the model is sufficiently worth when the model uses the boundary conditions in terms of the practical information of real plant instruments, where the system is small-scale and includes homogeneous fluid.
• But, the application of the model to the non-homogeneous system, such as the reactor coolant system, is difficult because of complexity of the modeling.

3. Institute-Academia Collaboration for Manpower Training
We have carried out manpower training program for nuclear engineering majored students through institute-academia collaborative lectures between KAERI and Chosun university, advisory tutoring of dispatched master course students at KAERI, and an advising graduate students for theses.
- Lecture opening for two semesters at the Chosun university for the institute-academia collaboration. The subjects include:
• Failure Analysis and Diagnostics (2nd semester, 2013)
• Evaluation of Equipment Integrity (1st semester, 2014)

- We have directed total three master course students who were dispatched for few months to KAERI. We have composed a paper as a result of the technical guidance on the common concern topic of the ‘NPP Early Diagnostic Research Center’, i.e., the instrument integrity in the case of severe accidents, and the collaborative researches between the first project and the second project.
- As a part of nuclear manpower training program, we have directed one master thesis and one doctorial thesis, respectively.

< KAIST >
1. Selection of safety-critical components which are needed to monitor and analysis of mechanical characteristics

Various safety-critical components which are needed to monitor in nuclear power plants were investigated, and safety-critical components which are needed to monitor were analyzed based on the investigation results. Furthermore, following results were obtained from the analysis of mechanical characteristics of selected components.

- Precise definition of terminologies used in nuclear power plants related to safety-critical components
- Investigation of safety-critical systems in nuclear power plants and investigation and analysis of included safety-critical components
- Selection of high pressure safety injection(HPSI) pump and motor operated valve which have high importances in nuclear power plants as safety-critical components which are needed to monitor among various safety-critical components
- Analysis of every failure causes and types of HPSI pump and motor operated valve
- Analysis of mechanical characteristic and tolerance of selected components

2. Instrumentation signal analysis for failure modes of selected components
Abnormal state factors of selected components were analyzed based on the selection criteria for safety-critical components which are needed to monitor. The results of figuring out instrumentation systems related to analyzed failure modes and investigation and analysis of possible abnormal signals are shown below.

- Investigation and analysis of abnormal signals from instrumentation systems according to analyzed failure modes based on investigation and analysis of instrumentation systems which are related to abnormal states of selected safety-critical components
- Analysis of abnormal state factors according to failure causes of selected components and evaluation of integrity and performance change according to temperature change

3. Development of integrity evaluation method for selected safety-critical components
Various methodologies were reviewed in order to develop integrity/state evaluation method for selected high pressure injection pump, and MFM methodology that many researches have been conducted in Europe, U.S., Japan, and China was selected among various methodologies. The results related to integrity/state evaluation method of HPSI pump using MFM model are shown below.

- Review of various prediction methodologies, selection of MFM methodology, and development of HPSI pump integrity/state evaluation method
• Since the HPSI pump is in standby state during normal operation, it is hard to acquire operation data and instrumentation systems which measure the performance of HPSI pump, are limited
• MFM methodology which is widely recognized as qualitative evaluation method was selected among artificial intelligence methodologies for HPSI pump integrity/state evaluation method.
• HPSI pump mainly operates during performance test and when safety injection signal(SIS) is actuated due to abnormality or accident occurrence in nuclear power plant
• In this project, modelling was performed for high pressure safety injection performance test parts
• Causes and consequences were estimated with MFM workbench under assuming abnormal states of HPSI pump as various cases. Estimated results were compared and analyzed with HPSI pump failure modes data

4. Human resource development
For professional manpower training in the field of nuclear industry and I&C, various training programs such as paperwork supervising, joint lectures, international workshops, and invited lectures given by foreign experts were conducted.

- Second KAIST international workshop was held. Contributed to professional manpower training with the chance to acquire expertise from presentation and discussion of foreign and domestic experts and exchange various information
- Lectures on theoretical backgrounds and practical training of MFM model were held with the invitation of professor LIND Morten from DTU(technical university of Denmark), who developed MFM
- Lectures on field works were held for graduate course with the invitation of researcher from KAERI(Korea atomic energy research institute)

( 출처 : SUMMARY 30p )

목차 Contents

• 표지 ... 1
• 제 출 문 ... 3
• 보고서 요약서 ... 5
• 요 약 문 ... 9
• SUMMARY ... 23
• CONTENTS ... 39
• 목차 ... 43
• 표목차 ... 45
• 그림목차 ... 49
• 제1장 연구개발과제의 개요 ... 55
• 제1절 연구개발의 목적 ... 55
• 제2절 연구개발 배경 및 필요성 ... 56
• 제3절 연구개발 범위 ... 59
• 제2장 국내외 기술개발 현황 ... 63
• 제1절 국내수준 ... 63
• 제2절 국내·외 기술개발 현황 ... 64
• 제3장 연구개발 수행내용 및 결과 ... 69
• 제1절 안전필수 계통, 설비의 상태 선제진단용 스마트 측정기술 개발 ... 69
• 1. 원전사고해석 데이터베이스 구축 ... 69
• 2. 상태 선제진단용 스마트센싱 기술 개발 ... 88
• 3. 밸브 및 배관 누설탐지기술 개발 ... 131
• 4. 원자력 인력양성을 위한 학연협력 ... 171
• 제2절 실시간 안전필수 계측기 건정성 진단기술 개발 ... 176
• 1. 실시간 안전필수 계측기 고장진단 방법론 및 모델 선정 ... 176
• 2. 실시간 안전필수계측기 고장진단 알고리즘 개발 및 구현 ... 220
• 3. 원자력 인력양성을 위한 학연협력 ... 283
• 제3절 실시간 원전 안전필수기기 건전성 평가 방법론 ... 288
• 1. 안전필수기기 조사 및 선정 ... 288
• 2. 선정된 기기의 고장모드 분석 ... 296
• 3. 고압안전주입펌프의 비정상 상태 인자 분석을 통한 계측기 신호 분석 ... 306
• 4. 안전필수기기 건전성 평가 방법론 ... 309
• 5. 선정된 방법론의 적용 ... 328
• 6. 원자력 인력양성 ... 346
• 제4장 목표달성도 및 관련분야에의 기여도 ... 349
• 제1절 목표달성도 ... 349
• 제2절 관련 분야 기여도 ... 352
• 1. 기술적 측면 ... 352
• 2. 경제적·산업적 측면 ... 354
• 3. 기술개발 내용의 공개 및 공유 ... 354
• 제5장 연구개발 결과의 활용계획 ... 355
• 제6장 연구개발과정에서 수집한 해외과학기술 정보 ... 357
• 제7장 참고문헌 ... 359
• 연구성과(연구사업지원시스템 입력성과) ... 365
• 끝페이지 ... 383