Loading...

Neutron Noise Reconstruction Based on Neuro-Fuzzy Computing for VVER-1000 Reactor Core

Pouyani Rad, Armin | 2018

93 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55314 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Vosoughi, Naser; Hosseini, Abolfazl
  7. Abstract:
  8. In this study, the reconstruction of the neutron noise source in the core of the VVER-1000 reactor was carried out using fuzzy neural network and parallel processing of parallel process circuit design. The noise of power reactors is derived from neutron fluctuations. Which are affected by fluctuations in reactor characteristics. These oscillations can be caused by the displacement of the elements forming the core or due to changes in temperature or density and so on. It is clear that any change in the elements of a reactor will make itself a change in the corresponding to cross-section of material. These changes called as pertubation. In this project, by using neuro-fuzzy network calculations, the noise source has been reconstructed on the basis of received data from neutron flux detectors inside the reactor core to determine the main cause of the pertubation. One of the problems in fuzzy neural computing is the speed of doing these calculations. Due to the use of MATLAB software toolbox and the lack of access to these codes, all of the nodes in the fuzzy neural network, as well as the relationships between them (neurons), have been programed to do their best, that they can be Optimized for existing problems. Also, this optimization can have a significant impact on software runtime. to increase computing speed is to use proprietary hardware to perform these calculations, because on existing computers, part of computer computing power is spent on operating system management, as well as running software and hardware connected to it. In fact, this processing part can be deleted and the entire processor power is used to perform the calculation. You can also leave yourself out of processor constraint and use proper processor and code optimization for computing. Outputs of this project include the coordinates of the noise source, the power and the frequency, and the hardware is suitable for ease of computing. Then, the design of the frequency section of circute was considered as the main challenge of designing a high-frequency parallel processing circuit, and after ensuring the correct operation of the high frequency part, the main hardware was designed with different needs and sectors. The hardware result was the ability to calculate 1000 billion logical instructions per second for every 3 watts consumed. Fuzzy neural network training has been done using simulated data in another study, of which 70% of these data are used for training the neural network and used for validation from another 30%. Finally, sensitivity analysis was performed on the number of detectors
  9. Keywords:
  10. Noise ; Reactor ; Neural Network ; Reconstruction ; Neuro Fuzzy Network ; Source Noise ; Bushehr Nuclear Power Plant ; Vodo-Vodyanoi Energetichesky (VVER)1000

 Digital Object List

 Bookmark

No TOC