Application of neural networks and neutron noise for diagnostics of reactor internals vibration

Conference · Sun Dec 31 04:00:00 EST 1995

OSTI ID:459144

Garis, N S; Pazsit, I ^[1]; Gloeckler, O ^[2]

---

1. Chalmers Univ. of Technology, Goeteborg (Sweden)
2. Ontario Hydro Nuclear, Toronto, Ontario (Canada)

It has long been known that vibration of reactor internals, in particular excessive vibrations of control rods, can be detected via the neutron noise they induce. Noise measurements are actually suitable to determine important diagnostic parameters such as the location of the vibrating rod and the vibration amplitude. An algorithm was earlier elaborated for this purpose, which is based on inversion of the expression describing the neutron noise as a function of vibration parameters. This inversion procedure is nevertheless complicated and not always unique. It was investigated whether a properly trained neural network can perform the inversion more effectively. It was found that artificial neural networks can be trained effectively to perform vibration diagnostics from neutron noise data fast, effectively and reliably. The present paper gives a description of the development and use of the neural networks for purposes of vibration diagnostics.

• OSTI ID: 459144
• Report Number(s): CONF-950420--
• Country of Publication: United States
• Language: English

Related Subjects

22 GENERAL STUDIES OF NUCLEAR REACTORS
CONTROL ELEMENTS
DIAGNOSTIC TECHNIQUES
DIAGNOSTIC USES
EXPERT SYSTEMS
MECHANICAL VIBRATIONS
NEURAL NETWORKS
REACTORS