PSAM 16 - Session Th12 Overview

PSAM 16 Conference Session Th12 Overview

Session Chair: Yong-Joon Choi (yong-joon.choi@inl.gov)

Paper 1 KR267

Lead Author: Heejong Yoo Co-author(s): Gyunyoung Heo (gheo@khu.ac.kr) *Corresponding author

Lessons-learned of using Monte Carlo method with importance sampling in fault tree quantification
In the quantitative evaluation of fault trees (FTs) and event trees (ETs) during a level 1 PSA, ETs are easily calculated by the product of each probability, while FTs need additional evaluation techniques to deal with Boolean logic. FTs used in nuclear engineering are usually classified as large FTs, which causes difficulty in calculating the top event probability when using the conventional methods. Other problems that could arise is that all conventional methods use minimal cut sets, which needs additional process of gaining the minimal cut sets. Validation issues are also present due to the fact that widely used methods are all using minimal cut sets, leading to the need of developing methods that is free of minimal cut sets. While there were some attempts to gain the top event probability of FTs by using the Monte Carlo method, which could be free from using minimal cut sets by setting a different algorithm, the time and computational cost for using the Monte Carlo method is always the main issue. In order to reduce computational resource and having its strong point in variance reduction, the most frequently used application for the Monte Carlo method is importance sampling. This paper suggests an algorithm of implying importance sampling, a general method used to reduce the cost for the Monte Carlo method, in order to quantify FTs, and show both the application and limitations of importance sampling. An example FT is given in the paper to show the application and algorithm of Monte Carlo method and to imply importance sampling for the quantification process.

Lessons-learned of using Monte Carlo method with importance sampling in fault tree quantification

In the quantitative evaluation of fault trees (FTs) and event trees (ETs) during a level 1 PSA, ETs are easily calculated by the product of each probability, while FTs need additional evaluation techniques to deal with Boolean logic. FTs used in nuclear engineering are usually classified as large FTs, which causes difficulty in calculating the top event probability when using the conventional methods. Other problems that could arise is that all conventional methods use minimal cut sets, which needs additional process of gaining the minimal cut sets. Validation issues are also present due to the fact that widely used methods are all using minimal cut sets, leading to the need of developing methods that is free of minimal cut sets. While there were some attempts to gain the top event probability of FTs by using the Monte Carlo method, which could be free from using minimal cut sets by setting a different algorithm, the time and computational cost for using the Monte Carlo method is always the main issue. In order to reduce computational resource and having its strong point in variance reduction, the most frequently used application for the Monte Carlo method is importance sampling. This paper suggests an algorithm of implying importance sampling, a general method used to reduce the cost for the Monte Carlo method, in order to quantify FTs, and show both the application and limitations of importance sampling. An example FT is given in the paper to show the application and algorithm of Monte Carlo method and to imply importance sampling for the quantification process.

Paper KR267 | Download the paper file. | Download the presentation PowerPoint file.

Name: Heejong Yoo (kreacher@khu.ac.kr)

Bio: Heejong Yoo, currently a Master's candidate, received the B.S. degree in Nuclear Engineering, Kyung Hee University, Gyeonggi-do, Republic of Korea, in 2021. His research interests are the multi-unit probabilistic safety assessments and cyber security for nuclear power plants. His recent activites include studies for the combination of site operation states in multi-unit PSA, analyzing PSA results with Monte Carlo method with importance sampling, emergency response in nuclear emergency situations, and cyber security for nuclear power plants.

Country: KOR
Company: Kyung Hee University
Job Title: Graduate Researcher

Paper 2 OH275

Lead Author: Kyusik Oh Co-author(s): Sangjun Park (sangjun@kaeri.re.kr) Gyunyoung Heo (gheo@khu.ac.kr) : corresponding author

Improving Measurement Reliability using Data Reconciliation and Digital Twin
The measurements may be less accurate because of defects in the measuring instruments or leakages in the system, but also of unknown reasons. These uncertainties are inevitable but can be managed if quantified. Error that can determine the cause, such as defects in the measuring instrument or leakages of facilities, is called gross error, and error that does not know the cause is called random error. In order to reduce or eliminate these errors, the application of data reconciliation and gross error detection technique is effective. These not only reduce random errors in measurements and eliminate gross errors, but also adjust the value to satisfy the correlations between them called physical models. These techniques were developed in various fields, but It seems rare that both how to update the physical model of the system and strategy to maintain the actual measurement system are explained simultaneously. First, this paper introduces in-house code (R language) using data reconciliation and gross error detection algorithms. Three case studies using power plant simulation data are explained. The second is the application using code, through the reconciled values it is identified whether the cause of the gross error is due to a defect in the instruments or an inaccurate physical models. As a result, it shows that the performance of maintenance increases using more accurate measurements and it can be an indicator of updating the physical models by verifying the accuracy of them.

Improving Measurement Reliability using Data Reconciliation and Digital Twin

The measurements may be less accurate because of defects in the measuring instruments or leakages in the system, but also of unknown reasons. These uncertainties are inevitable but can be managed if quantified. Error that can determine the cause, such as defects in the measuring instrument or leakages of facilities, is called gross error, and error that does not know the cause is called random error. In order to reduce or eliminate these errors, the application of data reconciliation and gross error detection technique is effective. These not only reduce random errors in measurements and eliminate gross errors, but also adjust the value to satisfy the correlations between them called physical models. These techniques were developed in various fields, but It seems rare that both how to update the physical model of the system and strategy to maintain the actual measurement system are explained simultaneously. First, this paper introduces in-house code (R language) using data reconciliation and gross error detection algorithms. Three case studies using power plant simulation data are explained. The second is the application using code, through the reconciled values it is identified whether the cause of the gross error is due to a defect in the instruments or an inaccurate physical models. As a result, it shows that the performance of maintenance increases using more accurate measurements and it can be an indicator of updating the physical models by verifying the accuracy of them.

Paper OH275 | | Download the presentation PowerPoint file.

Name: Kyusik Oh (ohsik95@gmail.com)

Bio: Kyusik Oh currently working in Kyung Hee University for M.S. degree. His research interests are data reconciliation in NPP facilities. Recently, he did research to improve the reliability of various systems, including research reactors, nuclear material processing processes, and turbine power plants.

Country: KOR
Company: Kyung Hee University
Job Title: M.S. degree

Paper 3 WO177

Lead Author: Woo Sik Jung Co-author(s): Seong Kyu Park (sparkpsa@ness.re.kr)

Balanced Fault Tree Modeling of Alternating Operating Systems in Probabilistic Safety Assessment
Nuclear power plants (NPPs) have alternating operation systems, such as the component cooling water system (CCWS), essential service water system (ESWS), essential chilled water system (ECWS), and chemical and volume control system (CVCS). Single-unit Probabilistic safety assessment (SUPSA) models for nuclear power plants (NPPs) have many failures of alternating systems. Furthermore, since NPPs undergo alternating operations between full power and low power and shutdown (LPSD), multi-unit PSA (MUPSA) models have failures of NPPs that undergo alternating operations between full power and LPSD. Their failures for alternating operations are modeled using fraction or partitioning events in seismic SUPSA and MUPSA fault trees. Since partitioning events for one system are mutually exclusive, their combinations should be excluded in exact solutions. However, it is difficult to eliminate the combinations of mutually exclusive events without modifying PSA tools for generating MCSs from a fault tree. If the combinations of mutually exclusive events are not deleted, core damage frequency (CDF) is underestimated. To avoid CDF underestimation in SUPSAs and MUPSAs, this paper introduces a process of converting partitioning events into conditional events, and conditional events are then inserted explicitly inside a fault tree. With this conversion, accurate CDF can be calculated without modifying PSA tools. It is strongly recommended that the method in this paper be employed for avoiding CDF underestimation in seismic SUPSAs and MUPSAs.

Balanced Fault Tree Modeling of Alternating Operating Systems in Probabilistic Safety Assessment

Nuclear power plants (NPPs) have alternating operation systems, such as the component cooling water system (CCWS), essential service water system (ESWS), essential chilled water system (ECWS), and chemical and volume control system (CVCS). Single-unit Probabilistic safety assessment (SUPSA) models for nuclear power plants (NPPs) have many failures of alternating systems. Furthermore, since NPPs undergo alternating operations between full power and low power and shutdown (LPSD), multi-unit PSA (MUPSA) models have failures of NPPs that undergo alternating operations between full power and LPSD. Their failures for alternating operations are modeled using fraction or partitioning events in seismic SUPSA and MUPSA fault trees. Since partitioning events for one system are mutually exclusive, their combinations should be excluded in exact solutions. However, it is difficult to eliminate the combinations of mutually exclusive events without modifying PSA tools for generating MCSs from a fault tree. If the combinations of mutually exclusive events are not deleted, core damage frequency (CDF) is underestimated. To avoid CDF underestimation in SUPSAs and MUPSAs, this paper introduces a process of converting partitioning events into conditional events, and conditional events are then inserted explicitly inside a fault tree. With this conversion, accurate CDF can be calculated without modifying PSA tools. It is strongly recommended that the method in this paper be employed for avoiding CDF underestimation in seismic SUPSAs and MUPSAs.

Paper WO177 | Download the paper file. | Download the presentation PowerPoint file.

Name: Woo Sik Jung (woosjung@sejong.ac.kr)

Bio: (1) Master and Doctoral Degrees for PSA at Korea Advanced Institute of Science and Technology, (2) Worked for KEPCO E&C, KAERI, and US EPRI, (3) Teaching Nuclear Engineering at Sejong University, (4) Developed fault tree solvers FORTE and FTREX that are interfaced with many PSA tools such as US EPRI tools (CAFTA, EOOS, FRANX, PHOENIX), SAPHIRE, SAREX, and AIMS-PSA

Country: KOR
Company: Sejong University
Job Title: Professor