Loading...

Source Localization by Analysis the Response of Detectors Using Inverse Methods

Mechershavi, Fatemeh | 2014

720 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 45469 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Vosoughi, Naser
  7. Abstract:
  8. Localization of a neutron point source using a designed computer program namely “MCMC-MATURE” is performed. The computer program analyses several detector responses in some certain media by Markov Chain Monte Carlo (MCMC) method and a new iteration algorithm. Identification of the possible regions of source position would be found by analyzing the initial fluxes generated by mesh tally of MCNPX computer code. The designed computer program is capable to generate the flux between detectors. “Regular-Sequential”, “Irregular-Sequential” and “Non-Sequential” are three methods used for sampling the generated random number in two dimensions. Each sample multiplied by a sampling function and lead to a random source position to make a location chain. The main assumed functions are uniform, sinusoidal, exponential and proportional which make the sampling function in three structures namely “simple”, “symmetric” and “asymmetric”. The acceptance or rejection of generated random locations is based on the Metropolis hasting algorithm. The Metropolis hasting ratio will check the generated random positions by analyzing their importance. This ratio ensures that the chain tends to the main position of source whenever its length increases. In iteration algorithm, the medium was meshed and each chain would increase a count in the proportional mesh. The number of counting which are saved in a matrix namely “probability counting matrix” would be normalized after investigation of a group chain. Investigating the next group chain will generate the other probability counting matrix. This probability counting matrix generating would be continued until the difference between the probability counting matrix and its previous one lead to a minimum quantity. This minimum quantity is determined for localization with notice to the maximum likelihood error and power analysis of the computer. The reasonable value of the minimum quantity in one and two dimensions is derived as 0.05 and 0.02 respectively.
    The sampling methods are investigated in two-dimensional uniform and Non-uniform media including water and wood. The best sampling method is obtained by Non-sequential sampling method. Amongst the sampler function, the exponential function with symmetric structure offers the best accuracy. Neutron source localization in different media using Non-Sequential method, exponential symmetric sampler function, 40 initial fluxes in each dimension, 2 flux interpolation, 81 meshes in each dimensions and iteration algorithm with 100 group chain is performed. In all investigations, the most likely position of the neutron source presence obtained by assuming the maximum error of the detector equal to 0.1, one flux interpolation and using uniform sampling function. In the dimensions of two meters, the mean error of accuracy and precision obtained 0.02 (cm) and 0.20% respectively and in two dimensional media obtained 0.05 (cm) and 0.38% respectively. The maximum error of accuracy in one and two dimensional media is 0.40% and 0.53% respectively.
    Sensitivity analysis of mentioned algorithm is investigated for several factors namely number of meshes, number of group chain and error of detector. The results show that the increase in number of meshes or number of group chain and decrease in maximum error of detector lead to less precision error and make the results more accurate
  9. Keywords:
  10. Location ; CF 252 Neutron Source ; Markov Chain Approximation Method ; Sampling ; Sampling Methods

 Digital Object List

 Bookmark

No TOC