DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Chang, Soon-Heung | - |
dc.contributor.advisor | 장순흥 | - |
dc.contributor.author | Lee, Gyu-Cheon | - |
dc.contributor.author | 이규천 | - |
dc.date.accessioned | 2011-12-14T08:05:18Z | - |
dc.date.available | 2011-12-14T08:05:18Z | - |
dc.date.issued | 2003 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=181066&flag=dissertation | - |
dc.identifier.uri | http://hdl.handle.net/10203/48947 | - |
dc.description | 학위논문(박사) - 한국과학기술원 : 원자력및양자공학과, 2003.2, [ xi, 107 p. ] | - |
dc.description.abstract | Improved methods of axial flux shape generation and DNBR calculation for designing an advanced digital core protection system in pressurized water reactors are presented in this study. The uncertainty of axial flux shape is relatively high since the coupling coefficients between axial flux shape and 3-level ex-core detector signals are based on the limited information obtained from startup tests. In addition, much conservatism is included in DNBR for compensating relatively long calculation time of complex sub-channel analysis algorithm. New methodologies using artificial neural networks are proposed to overcome these drawbacks. A feedforward network trained by backpropagation is used for axial flux shape generation. 20-node axial flux shapes are generated based on the three level ex-core detector signals. The ANN trained with 200 data predicts the 7,173 testing data with the average root mean square error of about 3 %. This method is also tested using the real plant data measured during normal operation. The RMS errors are within the range of 0.9-2.1%, which is about two times superior to the cubic spline approximation method currently used in the plant. This would result in the reduction of uncertainty in DNBR calculation, thereby the increase of the available thermal margin. Consequently, the developed AFS generation method would contribute to solve the drawback of the current method since it shows a reasonable accuracy over wide range of core conditions. A DNBR calculator has been also established using a radial basis function network and a wavelet neural network. These networks approximate the target DNBR by weighted sum of basis functions that nonlinearly squash the inputs. These networks provide more proper design methods for digital core protection systems than the traditional neural networks, since they have more concrete mathematical backgrounds. Nonparametric training approaches with these advanced networks showed dramatic reduction of the training t... | eng |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Axial Flux Shape | - |
dc.subject | Digital Core Protection System | - |
dc.subject | Artificial Neural Networks | - |
dc.subject | DNBR Calculation | - |
dc.subject | 핵비등이탈률 계산 | - |
dc.subject | 축방향 출력분포 | - |
dc.subject | 디지털 노심보호계통 | - |
dc.subject | 인공신경회로망 | - |
dc.title | Advanced digital core protection system using artificial neural networks | - |
dc.title.alternative | 인공신경회로망을 이용한 진보된 디지털 노심보호계통 연구 | - |
dc.type | Thesis(Ph.D) | - |
dc.identifier.CNRN | 181066/325007 | - |
dc.description.department | 한국과학기술원 : 원자력및양자공학과, | - |
dc.identifier.uid | 000965266 | - |
dc.contributor.localauthor | Lee, Gyu-Cheon | - |
dc.contributor.localauthor | 이규천 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.