Pipeline two-phase flow pressure drop algorithm for multiple inclinations

A Generalized Additive Model (GAM) is used to predict the pressure drop in two-phase flow at different inclinations and angles. The nonparametric nature of the method lets it have a high prediction capacity, but also has a great degree of interpretability due to the possibility to visualize the marginal effect of each predictor, unlike other machine learning methods. Also, the use of dimensionless numbers as predictors has a generalizability appeal. The GAM shows an outstanding capacity to predict the pressure gradient, having 99.1% adjusted R^2 and a mean relative error of 12.93%. This is even while ignoring bubbly flow in the training sample. A regularization double penalty approach was used, but most of the predictors are necessary to maintain the high predictive ability of the GAM. The model performs adequately on new data points not used on the training of the model randomly sampling the database. The splines and p-values of each term are shown, which help interpret the importance of the variables and their relationships with the pressure gradient.

• Autores: Cepeda Vega, Andrés Felipe
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2019
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/45118
• Acceso en línea: http://hdl.handle.net/1992/45118
• Palabra clave: Flujo bifásico; Modelos log-lineales; Tuberías; Fluidización; Ingeniería
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/