Loading...

Development of an Optimal Technique to Construct the Energy Spectrum of the Hpge Detector Using the Output Spectrum of the Nai Detector, with the Help of Soft Computing Algorithms

Yaghoubi Razgi, Zahra | 2022

63 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55831 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Vosoughi, Naser
  7. Abstract:
  8. Gamma ray spectroscopy has a special place in the industrial applications of nuclear radiation. Currently, the most common device for gamma ray detection and spectroscopy is the sodium iodide scintillation detector. Long life and high efficiency and reasonable price of these detectors are the reasons for the development of the use of these detectors in industries and laboratories. But these detectors in the classification of energy sensitive detectors are considered as detectors with low resolution. The presence of broad peaks in the gamma ray spectrum of this detector increases the possibility of interference of peaks related to different energies and makes it difficult to identify the spectrum. On the other hand, high pure germanium semiconductor detectors, despite their disadvantages in terms of maintenance, operating conditions and price, have excellent energy resolution. In this research, an attempt has been made to convert the output spectrum of the sodium iodide detector into the high pure germanium detector spectrum using machine learning techniques. The two techniques used for this mapping are neural network and linear regression. Learning from both methods is based on laboratory data. In the neural network approach, LSTM neural network is used in the structure of an AutoEncoder; In which, part of the spectra recorded in the laboratory are used in the form of 1024-long numerical sequences for training the network. The number of spectra used in the training process has been increased by adding random noise to the spectra recorded in the laboratory. Finally, the unused part of the laboratory spectra has been used to measure the correctness of the model's performance and its performance has been confirmed. In the regression approach, multiple regression is used with the help of the RandomForestRegressor estimator. A part of the recorded laboratory spectra was used for model training and another part was used for evaluation. In general, the spectra made by the neural network method are more accurate than the spectra made by the regression method with the RandomForestRegressor estimator
  9. Keywords:
  10. Machine Learning ; High Purity Germanium ; Gamma Ray Spectrometry ; Sodium Iodide Detector ; Energy Resolution ; Neural Network ; Linear Regression

 Digital Object List

 Bookmark

No TOC