The role of neural networks in reactor diagnostics and control

Journal Article · Sat Mar 01 04:00:00 EST 1997 · Advances in Nuclear Science and Technology

OSTI ID:447504

Pazsit, I ^[1]; Kitamura, M ^[2]

---

1. Chalmers Univ. of Technology, Goeteborg (Sweden)
2. Tohoku Univ., Sendai (Japan)

Reactor diagnostics and core diagnostics in particular, is an inverse task, just as most other diagnostics. One measures some physical parameter at some position, or the fluctuation thereof, which is given rise by the fluctuation of another parameter, presumably at a different position. In neutron noise diagnostics, the measured quantity is the neutron noise, whereas the cause, fluctuations of the core material, is called the {open_quotes}noise source{close_quotes}. The relationship between the cause (noise source) and the induced noise (effect) is determined by the physics of the process and can usually be described by a theory. This means that the direct task, calculation of the noise from the noise source, can always be achieved. In diagnostics, however, the process starts from the back end, i.e. one observes the effect of some cause. The task is to infer the cause (noise source) from the effect (induced noise), which is an inverse task (sometimes also called unfolding). The situation is very much similar even regarding control problems. The system state is described by a vector in the parameter space. When it deviates from the desired one, the control parameters need to be changed to bring the system into the desired state. The computation of the change in the state vector, due to a change in the control vector, can be calculated with no difficulty in principle. The control requires however the inverse task to be performed, i.e. to determine which changes of the control vector would bring the state vector into the desired state.

• OSTI ID: 447504
• Journal Information: Advances in Nuclear Science and Technology, Journal Name: Advances in Nuclear Science and Technology Vol. 24; ISSN 0065-2989; ISSN ANUTAC
• Country of Publication: United States
• Language: English

Related Subjects

22 GENERAL STUDIES OF NUCLEAR REACTORS
DIAGNOSTIC USES
NEURAL NETWORKS
REACTOR CONTROL SYSTEMS