Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Prediction of Soil-Water Characteristic Curve for Low-Plasticity Soils Using Multi-Output Neural Network
https://orcid.org/http://orcid.org/0000-0002-2744-5814
The soil water characteristic curve (SWCC) is considered the most used tool in describing the complex mechanical and hydrological behavior of unsaturated soils. In this paper, a dataset was collected from the Soil Survey Geographic Database. Utilizing this dataset and several machine learning techniques, various models were developed to accurately predict the volumetric water content at various levels of matric suction, including ensemble trees, support vector machines, and artificial neural networks (ANN). The predictive models were developed using 15 input variables representing different soil characteristics. The results indicated that the ANN models with multiple outputs outperformed other models in terms of coefficient of determination (R2), root mean square error, and mean absolute error with values of 0.99, 0.4869%, and 0.20%, respectively. Moreover, variable importance analysis was conducted using the best-performing ANN model. It was concluded that the most significant variables were the percent passing sieve number #200, cation exchange capacity, liquid limit, passing sieve number #10, and the plasticity index. Finally, training the best ANN was repeated using only the most significant input variables. A family of SWCC curves was developed using the best-performing ANN model with varying input parameters.
Prediction of Soil-Water Characteristic Curve for Low-Plasticity Soils Using Multi-Output Neural Network
https://orcid.org/http://orcid.org/0000-0002-2744-5814
The soil water characteristic curve (SWCC) is considered the most used tool in describing the complex mechanical and hydrological behavior of unsaturated soils. In this paper, a dataset was collected from the Soil Survey Geographic Database. Utilizing this dataset and several machine learning techniques, various models were developed to accurately predict the volumetric water content at various levels of matric suction, including ensemble trees, support vector machines, and artificial neural networks (ANN). The predictive models were developed using 15 input variables representing different soil characteristics. The results indicated that the ANN models with multiple outputs outperformed other models in terms of coefficient of determination (R2), root mean square error, and mean absolute error with values of 0.99, 0.4869%, and 0.20%, respectively. Moreover, variable importance analysis was conducted using the best-performing ANN model. It was concluded that the most significant variables were the percent passing sieve number #200, cation exchange capacity, liquid limit, passing sieve number #10, and the plasticity index. Finally, training the best ANN was repeated using only the most significant input variables. A family of SWCC curves was developed using the best-performing ANN model with varying input parameters.
Prediction of Soil-Water Characteristic Curve for Low-Plasticity Soils Using Multi-Output Neural Network
https://orcid.org/http://orcid.org/0000-0002-2744-5814
The soil water characteristic curve (SWCC) is considered the most used tool in describing the complex mechanical and hydrological behavior of unsaturated soils. In this paper, a dataset was collected from the Soil Survey Geographic Database. Utilizing this dataset and several machine learning techniques, various models were developed to accurately predict the volumetric water content at various levels of matric suction, including ensemble trees, support vector machines, and artificial neural networks (ANN). The predictive models were developed using 15 input variables representing different soil characteristics. The results indicated that the ANN models with multiple outputs outperformed other models in terms of coefficient of determination (R2), root mean square error, and mean absolute error with values of 0.99, 0.4869%, and 0.20%, respectively. Moreover, variable importance analysis was conducted using the best-performing ANN model. It was concluded that the most significant variables were the percent passing sieve number #200, cation exchange capacity, liquid limit, passing sieve number #10, and the plasticity index. Finally, training the best ANN was repeated using only the most significant input variables. A family of SWCC curves was developed using the best-performing ANN model with varying input parameters.
Prediction of Soil-Water Characteristic Curve for Low-Plasticity Soils Using Multi-Output Neural Network
Transp. Infrastruct. Geotech.
Mostafa, Omar (Autor:in) / Arab, Mohamed G. (Autor:in) / Hamad, Khaled (Autor:in)
Transportation Infrastructure Geotechnology ; 11 ; 4381-4404
01.12.2024
24 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Prediction of soil-water characteristic curve using neural network
| British Library Conference Proceedings | 2011
Prediction of Soil-Water Characteristic Curve Using Artificial Neural Network Approach
| British Library Conference Proceedings | 2018
Soils with Bimodal Soil-Water Characteristic Curve
| British Library Conference Proceedings | 2018
Soil-Water Characteristic Curve of Remolded Expansive Soils
| British Library Conference Proceedings | 2006