Thesis Estimación de pérdidas de hidrógeno por permeabilidad en ductos poliméricos
Loading...
Date
2025-03
Authors
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Química
Campus
Campus Casa Central Valparaíso
Abstract
La alta permeabilidad del hidrógeno en materiales poliméricos, como el polietileno de alta densidad (HDPE), representan un desafío crítico para garantizar la seguridad, eficiencia y sostenibilidad en el transporte de este vector energético. Es por lo anterior, que en este trabajo se cuantifican y modelan las pérdidas de hidrógeno por permeación en ductos de HDPE mediante la integración de herramientas computacionales avanzadas.
En primer lugar, se desarrolla y evalúa dos enfoques: un modelo matemático y un modelo basado en redes neuronales, para predecir el flujo de pérdida de hidrógeno en condiciones isotérmicas. Las predicciones consideraron variables operativas como la temperatura, el caudal, la presión de entrada y de salida, así como los parámetros de diseño del ducto (espesor, longitud, diámetro y elevación). Estos modelos se entrenan con los resultados de la simulación realizada a través de Aspen HYSYS, utilizando integración numérica del flujo de hidrogeno por unidad de área. Lo anterior es función del coeficiente de permeabilidad, el cual es modelado en función de la temperatura y la presión, basado en datos bibliográficos. Estos resultados se contrastan con un modelo más riguroso implementado en MATLAB, para evaluar la precisión de los resultados obtenidos en Aspen HYSYS.
El modelo para el coeficiente de permeabilidad demostró una alta precisión en función de la presión, con un error cuadrático medio (RMSE) de 0.078 y un coeficiente de correlación (R) de 0.999. Por otra parte, se obtuvo un RMSE de 0.149 y un 𝑅 de 0.999999 en el contraste entre Aspen HYSYS y el modelo implementado en MATLAB mostrando que los datos son representativos.
Finalmente, el modelo matemático logró predecir el flujo de pérdida de hidrógeno con un error menor al 5% en 24591 condiciones y el modelo de redes neuronales, con dos capas ocultas de 128 neuronas cada una, alcanzó predicciones con errores inferiores al 6% en el 99.5% de los casos y menores al 5% para pérdidas mayores a 20 mol/h, sin requerir el cálculo del factor de compresibilidad (Z) y el factor de fricción (𝑓).
The high permeability of hydrogen in polymeric materials, such as high-density polyethylene (HDPE), represents a critical challenge to ensure the safety, efficiency and sustainability in the transport of this energy vector. It is for this reason that this work quantifies and models hydrogen losses due to permeation in HDPE pipelines through the integration of advanced computational tools. First, two approaches are developed and evaluated: a mathematical model and a neural network-based model, to predict the flow of hydrogen loss under isothermal conditions. The predictions considered operating variables such as temperature, flow rate, inlet and outlet pressure, as well as pipeline design parameters (thickness, length, diameter, and elevation). These models are trained with the results of the simulation carried out through Aspen HYSYS, using numerical integration of the hydrogen flow per unit area. This is a function of the permeability coefficient, which is modeled as a function of temperature and pressure, based on bibliographic data. These results are contrasted with a more rigorous model implemented in MATLAB, to evaluate the accuracy of the results obtained in Aspen HYSYS. The model for the permeability coefficient demonstrated high accuracy as a function of pressure, with a mean square error (RMSE) of 0.078 and a correlation coefficient (R) of 0.999. On the other hand, an RMSE of 0.149 and 0.999999 was obtained 𝑅 in the Aspen HYSYS contrast and the model implemented in MATLAB showing that the data are representative. Finally, the mathematical model successfully predicted hydrogen loss flow with an error below 5% in 24,591 conditions, while the neural network model, with two hidden layers of 128 neurons each, achieved predictions with errors below 6% in 99.5% of cases and below 5% for losses greater than 20 mol/h, without requiring the calculation of the compressibility factor (Z) and the friction factor (𝑓).
The high permeability of hydrogen in polymeric materials, such as high-density polyethylene (HDPE), represents a critical challenge to ensure the safety, efficiency and sustainability in the transport of this energy vector. It is for this reason that this work quantifies and models hydrogen losses due to permeation in HDPE pipelines through the integration of advanced computational tools. First, two approaches are developed and evaluated: a mathematical model and a neural network-based model, to predict the flow of hydrogen loss under isothermal conditions. The predictions considered operating variables such as temperature, flow rate, inlet and outlet pressure, as well as pipeline design parameters (thickness, length, diameter, and elevation). These models are trained with the results of the simulation carried out through Aspen HYSYS, using numerical integration of the hydrogen flow per unit area. This is a function of the permeability coefficient, which is modeled as a function of temperature and pressure, based on bibliographic data. These results are contrasted with a more rigorous model implemented in MATLAB, to evaluate the accuracy of the results obtained in Aspen HYSYS. The model for the permeability coefficient demonstrated high accuracy as a function of pressure, with a mean square error (RMSE) of 0.078 and a correlation coefficient (R) of 0.999. On the other hand, an RMSE of 0.149 and 0.999999 was obtained 𝑅 in the Aspen HYSYS contrast and the model implemented in MATLAB showing that the data are representative. Finally, the mathematical model successfully predicted hydrogen loss flow with an error below 5% in 24,591 conditions, while the neural network model, with two hidden layers of 128 neurons each, achieved predictions with errors below 6% in 99.5% of cases and below 5% for losses greater than 20 mol/h, without requiring the calculation of the compressibility factor (Z) and the friction factor (𝑓).
Description
Keywords
Hidrógeno, Pérdidas por permeación, Simulación computacional, Polietileno de alta densidad
