The nature of statistical learning theory
The nature of statistical learning theory
Machine Learning
CIMCA '08 Proceedings of the 2008 International Conference on Computational Intelligence for Modelling Control & Automation
| Hi-index | 0.00 |
In a nuclear power plant (NPP), periodic sensor calibrations are required to assure sensors are operating correctly. However, only a few faulty sensors are found to be calibrated. For the safe operation of an NPP and the reduction of unnecessary calibration, on-line calibration monitoring is needed. Most researches have been focused on improving only the accuracy of the system although sensitivity is another important performance index. This paper presents multi-response optimization for an on-line sensor drift monitoring system to detect drift and estimate sensor signal effectively. Accuracy and sensitivity of the principal component-based auto-associative support vector regression (PCSVR) were optimized at the same time by desirability function approach. Response surface methodology (RSM) is employed to efficiently determine the optimal values of SVR hyperparameters. The proposed optimization method was confirmed with actual plant data of Kori NPP Unit 3. The results show the trade-off between the accuracy and sensitivity of the model as we expected.