@article { author = {Tiwari, Nand and Luxmi, KM and Ranjan, Subodh}, title = {Predictive Models for Prediction of Broad Crested Gabion Weir Aeration Performance}, journal = {Journal of Soft Computing in Civil Engineering}, volume = {7}, number = {2}, pages = {43-73}, year = {2023}, publisher = {Pouyan Press}, issn = {2588-2872}, eissn = {2588-2872}, doi = {10.22115/scce.2023.357761.1516}, abstract = {The gabion weirs serve the same functions that their counterpart impervious weirs do. However, they have the advantage of being eco-friendly, more stable, and economical in low to medium-head cases. Dissolved oxygen is one of the major determinants for the assessment of the purity of water. The purpose of the present work is to illustrate the comparison of multiple linear regression (MLR), neural network (NN), neuro-fuzzy system (NFS), deep neural network (DNN), and reported empirical models for the prediction of gabion weir aeration performance efficiency (APE20) with experimental results which are collected from the laboratory test. The NFS with four shaped membership functions, NN, DNN, MLR, and existing empirical models, are generated with the same input parameters, and their potentials are assessed to statistical appraisal indices. The results show that the DNN with the highest value of R2 (0.935) and NSE (0.934) and having the least errors in validating phase is the outperforming proposed model in the prediction of the APE20, which the NN model follows with R2 (0.917) and NSE (0.917). However, except trapezoidal shaped NFS model with R2 (0.873) and NSE (0.852) and MLR with R2 (0.905) and NSE (0.897), the remaining models of NFS-based and empirical relations could not perform better in validating phase. The sensitivity performance test is too conducted to find the relative relevance of the input parameter on the results of the APE20, where discharge per unit width (q) is found to be the most significant parameter, followed by the drop height (H0).}, keywords = {Gabion weir aeration- performance efficiency,Neural Network,Neuro-Fuzzy,Deep Neural Network,Empirical relations,Sensitivity performance}, url = {https://www.jsoftcivil.com/article_168919.html}, eprint = {https://www.jsoftcivil.com/article_168919_3a0d5e158125a7de9349aa7bc6c69a1c.pdf} }