Item


Modeling pressure drop produced by different filtering media in microirrigation sand filters using the hybrid ABC-MARS-based approach, MLP neural network and M5 model tree

Granular media filters are commonly used to remove suspended solids and prevent emitter clogging in microirrigation systems. Silica sand is the standard filtering media but other granular materials can be used for this purpose. The characterization of the pressure drop produced by the clean filtering media is of practical interest for designing and managing these filters. Different models such as Ergun or Kozeny-Carman equations are usually used to predict pressure drop produced by the filtering media. However, as parameters of the media such as equivalent diameter and sphericity, that are difficult to determine, appear in these equations, the objective of this study was to construct a new model to estimate the pressure drop of different filtering materials of interest in granular filters with limited data of the physical parameters that characterize the filtering media. This paper, taking as starting point the multivariate adaptive regression splines (MARS), develops a new algorithm hybridizing it with the artificial bee colony (ABC) method, to estimate the pressure drop in granular filters broadly used in microirrigation systems for the first time with a data-driven model. Laboratory experiments were used to measure pressure drop across silica sand, crushed glass, modified glass and glass microspheres in a scaled filter at surface velocities ranging from 0.004 to 0.025 m s −1 . The ABC method allows the tuning of the MARS parameters during the training phase improving significantly the regression accuracy. Additionally, a multilayer perceptron network (MLP) and M5 model tree were fitted to the experimental data for comparison purposes. The results have shown that ABC-MARS-based model was the best estimation of the pressure drop with a coefficient of determination of 0.78. Therefore, ABC-MARS-based model could be easily implemented to predict pressure drop with minimal input parameters for other filtering materials used in microirrigation media filters

Authors wish to acknowledge the computational support provided by the Department of Mathematics at University of Oviedo as well as financial support of the Spanish Ministry of Economy and Competitiveness through Grant AGL2015-63750-R

© Computers and Electronics in Agriculture, 2017, vol. 139, p. 65-74

Elsevier

Author: García Nieto, P. J.
García-Gonzalo, E.
Bové Masmiquel, Josep
Arbat Pujolràs, Gerard
Duran i Ros, Miquel
Puig Bargués, Jaume
Date: 2017 June 15
Abstract: Granular media filters are commonly used to remove suspended solids and prevent emitter clogging in microirrigation systems. Silica sand is the standard filtering media but other granular materials can be used for this purpose. The characterization of the pressure drop produced by the clean filtering media is of practical interest for designing and managing these filters. Different models such as Ergun or Kozeny-Carman equations are usually used to predict pressure drop produced by the filtering media. However, as parameters of the media such as equivalent diameter and sphericity, that are difficult to determine, appear in these equations, the objective of this study was to construct a new model to estimate the pressure drop of different filtering materials of interest in granular filters with limited data of the physical parameters that characterize the filtering media. This paper, taking as starting point the multivariate adaptive regression splines (MARS), develops a new algorithm hybridizing it with the artificial bee colony (ABC) method, to estimate the pressure drop in granular filters broadly used in microirrigation systems for the first time with a data-driven model. Laboratory experiments were used to measure pressure drop across silica sand, crushed glass, modified glass and glass microspheres in a scaled filter at surface velocities ranging from 0.004 to 0.025 m s −1 . The ABC method allows the tuning of the MARS parameters during the training phase improving significantly the regression accuracy. Additionally, a multilayer perceptron network (MLP) and M5 model tree were fitted to the experimental data for comparison purposes. The results have shown that ABC-MARS-based model was the best estimation of the pressure drop with a coefficient of determination of 0.78. Therefore, ABC-MARS-based model could be easily implemented to predict pressure drop with minimal input parameters for other filtering materials used in microirrigation media filters
Authors wish to acknowledge the computational support provided by the Department of Mathematics at University of Oviedo as well as financial support of the Spanish Ministry of Economy and Competitiveness through Grant AGL2015-63750-R
Format: application/pdf
Citation: https://doi.org/10.1016/j.compag.2017.05.008
ISSN: 0168-1699
Document access: http://hdl.handle.net/10256/14465
Language: eng
Publisher: Elsevier
Collection: MINECO/PE 2016-2018/AGL2015-63750-R
Reproducció digital del document publicat a: https://doi.org/10.1016/j.compag.2017.05.008
Articles publicats (D-EQATA)
Is part of: © Computers and Electronics in Agriculture, 2017, vol. 139, p. 65-74
Rights: Tots els drets reservats
Subject: Regatge per degoteig
Trickle irrigation
Filtres i filtració
Filters and filtration
Title: Modeling pressure drop produced by different filtering media in microirrigation sand filters using the hybrid ABC-MARS-based approach, MLP neural network and M5 model tree
Type: info:eu-repo/semantics/article
Repository: DUGiDocs

Subjects

Authors