Modeling of discharge Coefficient of nonlinear weirs with QNET and SVM methods
In this study, we investigate the discharge coefficient prediction of arched labyrinth weirs with a cycle angle of 6 degrees, employing 243 data series. Labyrinth weirs, aside from their economic advantages, exhibit superior flow-passing capabilities compared to linear weirs. Notably, increasing the...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Shahid Chamran University of Ahvaz
2024-05-01
|
| Series: | Journal of Hydraulic Structures |
| Subjects: | |
| Online Access: | https://jhs.scu.ac.ir/article_19040_e0ffb62d5e2309b67d6291db54e4d8ae.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In this study, we investigate the discharge coefficient prediction of arched labyrinth weirs with a cycle angle of 6 degrees, employing 243 data series. Labyrinth weirs, aside from their economic advantages, exhibit superior flow-passing capabilities compared to linear weirs. Notably, increasing the length of the crest within a specified width enhances discharge efficiency with less upstream height. Machine learning algorithms, namely QNET and SVM, play a pivotal role due to their proficiency in uncovering intricate relationships between independent and dependent parameters, leading to significant time and cost savings. Our results using QNET and SVM indicate that the combination of (Cd, , θ, ) yields optimal accuracy, with QNET achieving (R²=0.9850), (RMSE=0.0259), and (DC=0.9892) in the training phase, and (R²=0.9824), (RMSE=0.0292), and (DC=0.9788) in the test phase. For SVM, the training phase results are (R²=0.9889), (RMSE=0.0189), and (DC=0.9870), and in the test phase (R²=0.9881), (RMSE=0.0199), and (DC=0.9853). Sensitivity analysis shows the significant role of the total water load ratio parameter ( ) in determining the discharge coefficient of arched labyrinth weirs. This research contributes to the understanding of non-linear arched weir discharge predictions and highlights the efficacy of QNET and SVM algorithms in this domain. |
|---|---|
| ISSN: | 2345-413X 2345-4156 |