Loading...

Kinetic modeling of gas hydrates formation in presence of additives using Artificial Neural Networks

Mehruz Monfared, Saeed | 2012

739 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43470 (06)
  4. University: Sharif University of Technology
  5. Department: Chemical and Petroleum Engineering
  6. Advisor(s): Taghikhani, Vahid; Ghotbi, Cyrus
  7. Abstract:
  8. Many researchers were interested in gas hydrates because of their especial properties. These researches divided in two fields, thermodynamic and kinetic. The studies on first field began since many years ago, and they provided suitable results. But the investigation of hydrate formation kinetic started for 30 years ago, approximately, and according to its features such as high complexity of crystallization process and severe dependency on system, yet the comprehensive and pervasive model was not presented. At this project, after the presentation and survey of diverse models, in order to kinetic modeling of available data, among different kinetic models a reaction model selected. Then primary neural network designed with the help of two different methods, and the designed network by means of growing method waschosen as a best network. After the selection of best network, optimization of network done. Then, transfer functions, subject function and backpropagation algorithm were changed. According to final tests on about 16000 neural networks, a network with two hidden layers, which has 12 neurons in first layer and 13 neurons in second layer, selected as a best network. Dimension reduction algorithms, k-order cross folding tests and Leave-one-out method were used to validate the network, and they validated proper inputs, network comprehensiveness and lack of overfitting. Finally, the results of neural network and that of reaction modelcompared. The results showed that the neural network error was 84% lower than reaction model error
  9. Keywords:
  10. Neural Network ; Gas Hydrate ; Network Optimization ; Hydrat Formation Kinetics

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC