|
" A method to predict liquid entrainment fraction and quantify the associated uncertainty in two-phase annular flow "
Md Azharul Islam
Crunkleton, Daniel
| Document Type
|
:
|
Latin Dissertation
|
| Language of Document
|
:
|
English
|
| Record Number
|
:
|
804356
|
| Doc. No
|
:
|
TL49185
|
| Call number
|
:
|
1864680736; 10250292
|
| Main Entry
|
:
|
Jebbouri, Abdessamia
|
| Title & Author
|
:
|
A method to predict liquid entrainment fraction and quantify the associated uncertainty in two-phase annular flow\ Md Azharul IslamCrunkleton, Daniel
|
| College
|
:
|
The University of Tulsa
|
| Date
|
:
|
2016
|
| Degree
|
:
|
M.S.
|
| field of study
|
:
|
Chemical Engineering
|
| student score
|
:
|
2016
|
| Page No
|
:
|
160
|
| Note
|
:
|
Committee members: Cremaschi, Selen; Ramsurn, Hema
|
| Note
|
:
|
Place of publication: United States, Ann Arbor; ISBN=978-1-369-43673-0
|
| Abstract
|
:
|
A methodology was used to predict the entrainment fraction in two- phase annular flow for a given input condition and the uncertainty associated with the prediction was calculated. The tested methodology was independent of inclination angle, pressure range and fluids used in the annular flow. For a given input condition, the applied methodology used a set of experimental data to train and evaluate 17 different liquid entrainment models and selected the best model based on the experimental data. The uncertainty of the prediction was calculated by propagating the Monte Carlo simulation method. A data validation method was used to evaluate the prediction performance of the tested methodology for an experimental database collected from the open literature. Data validation method showed that current study can predict 94% experimental data within ±10% error limit and the best available model can predict only 50% data within ±10% error limit. An Extensive statistical analysis was performed to evaluate the performances of 18 different liquid entrainment models and best performing models for different flow condition were identified. Euclidean distances were calculated from the input condition to experimental data to collect the relevant experimental data from the database for the training and the evaluation of models. In order to select the best model for a given input condition, models were screened and ranked by an extensive statistical analysis. Associated uncertainty of the prediction was calculated for the input condition and the experimental data. During the prediction of each input condition, semi-mechanistic models were fine-tuned with the relevant data for optimum performance. Based on the evaluations of 18 different models with 1711 experimental data set, the Mantilla (2008) model performed the best for the horizontal annular flow data. For the vertical annular flow data, the Oliemans et al. (1986) model was found to be better than any other models and for the inclined annular flow data, the Paleev and Filippovich (1966) model gave the best prediction accuracy.
|
| Subject
|
:
|
Chemical engineering
|
| Descriptor
|
:
|
Applied sciences;Liquid entrainment
|
| Added Entry
|
:
|
Crunkleton, Daniel
|
| Added Entry
|
:
|
Chemical EngineeringThe University of Tulsa
|
| |