Study on automatic ECT data evaluation by using neural network

Conference · Sat Dec 31 04:00:00 EST 1994

OSTI ID:72628

Komatsu, H; Matsumoto, Y; Badics, Z; Aoki, K; Nakayasu, F ^[1]; Hashimoto, M ^[2]; Miya, K ^[3]

---

1. Nuclear Fuel Industries Ltd., Osaka (Japan)
2. Univ. of Industrial Technology, Kanagawa (Japan)
3. Univ. of Tokyo (Japan)

At the in--service inspection of the steam generator (SG) tubings in Pressurized Water Reactor (PWR) plant, eddy current testing (ECT) has been widely used at each outage. At present, ECT data evaluation is mainly performed by ECT data analyst, therefore it has the following problems. Only ECT signal configuration on the impedance trajectory is used in the evaluation. It is an enormous time consuming process. The evaluation result is influenced by the ability and experience of the analyst. Especially, it is difficult to identify the true defect signal hidden in background signals such as lift--off noise and deposit signals. In this work, the authors performed the study on the possibility of the application of neural network to ECT data evaluation. It was demonstrated that the neural network proved to be effective to identify the nature of defect, by selecting several optimum input parameters to categorize the raw ECT signals.

• OSTI ID: 72628
• Report Number(s): CONF-931061--; ISBN 0-87170-506-0
• Country of Publication: United States
• Language: English

Related Subjects

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS
DATA ACQUISITION
DATA ANALYSIS
EDDY CURRENT TESTING
IN-SERVICE INSPECTION
NEURAL NETWORKS
PWR TYPE REACTORS
STEAM GENERATORS
TUBES