Hybrid diagnostic agent system for the fast-breeder reactor “Monju”

GOFUKU Akio¹, TAKAHASHI Makoto², NAGAMATSU Takashi³,
MOCHIZUKI Hiroyasu⁴, FURUSAWA Hiroaki¹, and MINOWA Hirotsugu¹

1. Graduate School of Natural Science and Technology, Okayama University, 3-1-1 Tsushima-Naka, Kita-ku, Okayama 700-8530, Japan (fukuchan@sys.okayama-u.ac.jp)
2. Graduate School of Engineering, Tohoku University, 6-6 Aramaki-Aoba, Aoba-ku, Sendai, Miyagi 980-8579, Japan
3. Graduate School of Maritime Sciences, Kobe University, 5-1-1 Fukaeminami-machi, Higashinada-ku, Kobe 658-0022, Japan
4. Research Institute of Nuclear Engineering, University of Fukui, 1-2-4 Kanawa-cho, Tsuruga 914-0055, Japan

Abstract: A hybrid diagnostic agent system is developed to detect and identify early an anomaly that might happen in the fast-breeder reactor “Monju”. The system outputs a diagnostic result by integrating the results of diagnosis by four diagnostic software agents. They are (1) an estimation agent of overall heat transfer coefficient of evaporator and superheater, (2) a state identification agent based on SVM (Support Vector Machine), (3) an anomaly detection agent by WT (Wavelet Transformation), and (4) a CBR (Case-Based Reasoning) agent using several attributes in both time and frequency domains. This article describes the four diagnostic techniques and the hybrid diagnostic agent system.
Keyword: diagnostic system; fast-breeder reactor; hybrid agent system