Pressure drop modelling in sand filters in micro-irrigation using gradient boosted regression trees
Pressure losses across filters should be known for the proper design and management of this irrigation equipment. Pressure losses produced by filtering media in sand filters can be computed using Ergun or Kozeny-Karman equations, which require knowledge, among other parameters, of the sphericity of...
| Autores: | , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2018 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:10256/24128 |
| Acceso en línea: | http://hdl.handle.net/10256/24128 |
| Access Level: | acceso abierto |
| Palabra clave: | Regatge per degoteig Trickle irrigation Filtres i filtració Filters and filtration |
| Sumario: | Pressure losses across filters should be known for the proper design and management of this irrigation equipment. Pressure losses produced by filtering media in sand filters can be computed using Ergun or Kozeny-Karman equations, which require knowledge, among other parameters, of the sphericity of the filter medium. As this parameter is not easy to determine, it is useful to explore the performance of alternative computing methods that can avoid requiring knowledge of sphericity. In this paper, taking as starting point the nonparametric machine learning approach known as the gradient boosted regression tree (GBRT) approach and hybridising it with the differential evolution (DE) technique, the pressure drop in sand filters used in microirrigation has been modelled. For different filtering materials such as modified glass, crushed glass, silica sand and glass microspheres, experimental data of pressure drop for velocities between 0.004 and 0.025 m s−1 was collected and the model built. The results demonstrated that DE-GBRT-based model was able to accurately predict pressure drop. The model also allowed ranking of the importance of the independent variables examined within the model. Taking into account this ranking, and using only the main variables, a simplified method with an improved coefficient of determination was constructed |
|---|