Estimating the scour depth of slope control structures with sharp-crested weir using artificial intelligence models
Abstract Introduction: In free overfall spillways, waterfalls over the crown of the spillway almost vertically and impacts the downstream bed of the dams. Due to the high velocity and energy of the flow which impacts the erodible downstream bed, it may cause scouring close to the foundation of the d...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | fas |
| Published: |
Marvdasht Branch, Islamic Azad University
2023-01-01
|
| Series: | مهندسی منابع آب |
| Subjects: | |
| Online Access: | https://wej.marvdasht.iau.ir/article_5809_58a0efca30bc4479772b7b5b3bca3283.pdf |
| Summary: | Abstract
Introduction: In free overfall spillways, waterfalls over the crown of the spillway almost vertically and impacts the downstream bed of the dams. Due to the high velocity and energy of the flow which impacts the erodible downstream bed, it may cause scouring close to the foundation of the dam and consequently threaten the stability of the dam.
Methods: In this study, artificial intelligence methods were used to estimate the scour depth of slope control structures with sharp-crested weir due to the complexity of the phenomenon. Three models including neural network, adaptive fuzzy neural system, and support vector machine (SVM) were used as artificial intelligence or black-box model to solve the problem..
Findings: The results showed that artificial intelligence methods are more efficient than conventional experimental methods in estimating the depth of downstream scours of slope control structures with sharp-crested weir. Using more parameters in the input of artificial intelligence models does not increase the accuracy of these models. It is because of increasing errors as a result of using more parameters in these models. In estimating the downstream scour depth of slope control structures with the sharp-crested weir in both calibration and validation stages, an adaptive fuzzy neural system model is up to 20% more reliable than the artificial neural network model and up to 8.5% than the support vector machine model. |
|---|---|
| ISSN: | 2008-6377 2423-7191 |