Evaluating Parshall flume aeration with experimental observations and advance soft computing techniques

• Published in: Neural computing & applications Vol. 33; no. 24; pp. 17257 - 17271
• Main Authors: Bhoria, Sangeeta, Sihag, Parveen, Singh, Balraj, Ebtehaj, Isa, Bonakdari, Hossein
• Format: Journal Article
• Language: English
• Published: London Springer London 01.12.2021; Springer Nature B.V
• Subjects: Aeration; Aquatic life; Artificial Intelligence; Canals; Civil engineering; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Construction; Data Mining and Knowledge Discovery; Discharge measurement; Efficiency; Flow velocity; Flumes; Froude number; Group method of data handling; Hydraulics; Image Processing and Computer Vision; Irrigation canals; Mars; Mean square errors; Open channels; Original Article; Probability and Statistics in Computer Science; Reynolds number; Root-mean-square errors; Seepage; Sewage treatment plants; Soft computing; Throats; Water measurement; Parshall flume; Generalized structure-group method of data handling; Multivariate adaptive regression splines; Aeration efficiency
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-021-06316-9

A Parshall flume is a versatile device due to its diverse applications in irrigation canals, mine discharge, dam seepage, sewage treatment plants and many more. It is an economical due to its construction and installation and an accurate in discharge measurement in open channel and non-full pipe. Exchange of oxygen between water and air is termed as aeration. In the study, advance soft computing models; Multivariate Adaptive Regression Splines (MARS) and Generalized Structure- Group Method of Data Handling (GS-GMDH) have used to predict aeration efficiency ( E 20 ) values at Parshall flumes and its modified forms and also compared with existing conventional models. The performance of the models was assessed using four evaluating metrics; coefficient of determination ( R 2 ), mean absolute error (MAE), root mean square error (RMSE) and Nash Sutcliffe model efficacy. Agreement plot of GS-GMDH and MARS models showed that the MARS model has the maximum exactness in predicting E 20 values at Parshall flumes and its modified forms with the lowest RMSE = 0.0020, MAE = 0.0015. MARS model with input combination of ratio of throat width to throat length (W/L), ratio of sill height to throat length (S/L) and Froude number is performing better than that of all other input combinations used for models and found to be more suitable for predicting the E 20 . Overall comparison among conventional and advance soft computing models suggests that advance soft computing models perform better than conventional models.