Modelación estadística de fallas en redes de distribución de agua potable : caso de estudio: Bogotá, Colombia
"The main objective of WDS is to supply water to the population in the required quantity and quality. Factors such as deterioration of system components, climate change, the growing water demand and economic restrictions increased the complexity of their management. Pipe failures in water distr...
- Autores:
-
Giraldo González, Mónica Marcela
- Tipo de recurso:
- Fecha de publicación:
- 2018
- Institución:
- Universidad de los Andes
- Repositorio:
- Séneca: repositorio Uniandes
- Idioma:
- spa
- OAI Identifier:
- oai:repositorio.uniandes.edu.co:1992/60934
- Acceso en línea:
- http://hdl.handle.net/1992/60934
- Palabra clave:
- Corrosión de tuberías
Minería de datos
- Rights
- openAccess
- License
- https://repositorio.uniandes.edu.co/static/pdf/aceptacion_uso_es.pdf
| Summary: | "The main objective of WDS is to supply water to the population in the required quantity and quality. Factors such as deterioration of system components, climate change, the growing water demand and economic restrictions increased the complexity of their management. Pipe failures in water distribution systems generate economic, environmental and social costs. These include high repair costs, interruption of the water's supply and traffic, contaminant intrusion into the network and environmental resources loss, including water and energy. The effective planning renovation of the WDS requires an accurate quantification of the pipe's structural deterioration. Direct inspection of these structures is often difficult and expensive. The implementation of statistical models and data-driven methods to predict pipe failures is an effective and economic alternative. In this investigation, several statistic and data-driven models are presented for more comprehensive and accurate prediction of pipe failure. To predict failures in pipe groups, three models, which relate the number of failures with the pipe?s attributes, are implemented: multiple linear regression, Poisson regression and EPR. Adjusted regressions showed acceptable results in terms of their performance (R2 between 0.695 and 0.927). The results demonstrate that simple models along with pipes physical variables are capable to estimate the number of failures in homogeneous groups. In the case of individual pipes, four models were employed (i.e, Gradient Boosted tree, Bayes, SVM and ANN). The pipe's attributes and some environmental and operational variables were included as input variables. The models performance evaluation reveals acceptable results except for the ANNs. The model that presented the best performance in terms of predictive capacity was the decision tree technique."--Tomado del Formato de Documento de Grado. |
|---|