Computation of Scour Depth Below Pipelines using Artificial Neural Networks Ansari Mujib Ahmad, Ansari Sarfaraz Ali, Alam Shahid Department of Civil Engineering, Z.H. College of Engineering & Technology, Aligarh Muslim University, India Online published on 22 April, 2019. Abstract The mechanism of flow generated around the pipeline and river bed is so complicated that it is difficult to establish a general regression model to accurately estimate the scour depth below pipe lines. Hence in this paper an alternative approach to the conventional regression approach in the form of ANNs is proposed to predict the scour depth below pipe lines. The experimental data collected from literature having a wide range of hydraulic and geometrical variables are used to train, test and validate the network. A network architecture complete with trained values of connection weight and bias and requiring input of ungrouped parameters pertaining to (ρ’s, y, D, d50, Sf, e, B, and V) is recommended in order to predict the scour depth below pipe lines. Predictions based on the original raw data (ρ’s, y, D, d50, Sf, e, B, and V) were better than those based on grouped dimensionless forms of data (τ*, y/D, D/d50, F, Re, Sf, and e/D). On the basis of sensitivity analysis, it is observed that the pipe diameter (D) is the most significant parameter. The variables in order of decreasing level of sensitivity for Model M1 with CFBP are: D,V, y, d50, e, Sf, B and ρ’s. It was found that ANNs results were better than those of regression model. Top Keywords Artificial neural network, regression, scour depth, pipe line, sediment. Top |