Extreme Learning Machines in Predicting the Velocity Distribution in Compound Narrow Channels

Due to the importance of velocity distribution in open channels, this matter has long been of interest to scientists and engineers. Studies show that the maximum velocity in open channels often occurs below the free surface, something known as the dip phenomenon. This phenomenon can serve as a measure in river hydraulics and sediment transport studies. However, existing relationships for estimating the dip phenomenon has advantages and disadvantages, with the biggest problem being poor performance in various hydraulic conditions. To overcome such problems and to achieve high modeling speed and good performance, Extreme Learning Machines (ELM), a new soft computing technique for velocity field modeling is used in this study. This method is a novel algorithm for feedforward neural networks and has only one hidden layer. Field-measured datasets from a narrow compound open channel in France are employed to model the velocity field. The results indicate that ELM performed adequately (MAPE = 3.03; RMSE = 0.03) in velocity field prediction. However, it was found that the ELM offers higher performance than the existing methods.

[1]  Xianzhong Long,et al.  Discriminative graph regularized extreme learning machine and its application to face recognition , 2015, Neurocomputing.

[2]  Vijay P. Singh,et al.  Two-Dimensional Velocity Distribution in Open Channels Using the Tsallis Entropy , 2013 .

[3]  Vijay P. Singh,et al.  Entropy Theory for Two-Dimensional Velocity Distribution , 2011 .

[4]  Hossein Bonakdari,et al.  Evaluation of Sediment Transport in Sewer using Artificial Neural Network , 2013 .

[5]  Naif Alajlan,et al.  Fusion of Extreme Learning Machine and Graph-Based Optimization Methods for Active Classification of Remote Sensing Images , 2015, IEEE Geoscience and Remote Sensing Letters.

[6]  Hossein Bonakdari,et al.  Performance Evaluation of Adaptive Neural Fuzzy Inference System for Sediment Transport in Sewers , 2014, Water Resources Management.

[7]  Amir Hossein Zaji,et al.  Adaptive neuro-fuzzy inference system multi-objective optimization using the genetic algorithm/singular value decomposition method for modelling the discharge coefficient in rectangular sharp-crested side weirs , 2016 .

[8]  Frédérique Larrarte,et al.  Turbulent velocity profile in fully-developed open channel flows , 2008, Environmental Fluid Mechanics.

[9]  Hossein Bonakdari,et al.  Modeling of velocity fields by the entropy concept in narrow open channels , 2015 .

[10]  Hossein Bonakdari,et al.  Assessment of evolutionary algorithms in predicting non-deposition sediment transport , 2015 .

[11]  Amir Hossein Zaji,et al.  Predicting the velocity field in a 90° Open channel bend using a gene expression programming model , 2015 .

[12]  Frédérique Larrarte,et al.  Turbulent velocity profiles : a new law for narrow channels , 2010 .

[13]  Özgür Kisi,et al.  Long-term monthly evapotranspiration modeling by several data-driven methods without climatic data , 2015, Comput. Electron. Agric..

[14]  Frédérique Larrarte,et al.  Profil de vitesses turbulent : une nouvelle loi pour les canaux étroits , 2010 .

[15]  Robert M. Boes,et al.  Turbulence Characteristics in Supercritical Open Channel Flows: Effects of Froude Number and Aspect Ratio , 2014 .

[16]  Chee Kheong Siew,et al.  Extreme learning machine: Theory and applications , 2006, Neurocomputing.

[17]  Qunxiong Zhu,et al.  A data-attribute-space-oriented double parallel (DASODP) structure for enhancing extreme learning machine: Applications to regression datasets , 2015, Eng. Appl. Artif. Intell..