Parametric Sensitivity Analysis in Local Scouring Depth of Bridge Piles using the Smart Model

Alizadeh, Marjan (2018) Parametric Sensitivity Analysis in Local Scouring Depth of Bridge Piles using the Smart Model. Masters thesis, University of Mohaghegh Ardabili.

Preview

Text (آنالیز حساسیت پارامترهای موثر بر عمق آب‌شستگی موضعی پایه پل با استفاده از مدل هوشمند) Marjan Alizadeh.pdf Download (1MB) | Preview

Abstract

One of the destructive factors in bridges which lie in river path is the scouring of the base of the bridge. Because of the unavailablity of the universal theorotic fundamental and complicated relation of the scouring depth with various parameters, the results of the different relations are acceptable for special conditions but they have high dispersity. In order to achieve the scouring depth values for each region, the comparison of the case with physical models has been done. Hence achieving a method that can present precisly results and the ability to extend the procedure to the other data range is so important. In this research the prediction of the bridge base is done using artificial neural network as calculating tools and then the sensitivity analysis has been done for every parameter that is used in the network. The mathematic neural network that is used in this research is the Multi Layer Percepteron (MLP) type that used the Back Propagation error (BP) for the network train process. Frthermore in order to decrease the calculation, the discrete wavelet analysis has been used in comparison to the continues one. The five effective input parameters are pier diameter, granuler diameter, current depth, current mean velocity and current critical velocity. The optimum neural network determined by investigation of different factorsperformances including the training law transfer function, the number of hidden layers and the number of nodes for each layer. The optimised network has been determined using these parameters and then the sensitivity of the network for each parameters has been probed individually. Considering the error and correlation statistical indice, the efficiency of the networks have been assessed using the mentioned method. Numerical results demonstrate the high ability of the neural network and the combined wavelet neural network in predicting of the scouring depth of the pier in comparison to the empirical formullas and method and the results of the sensitivity analysis depict that the pier diameter plays the key rol for precasting the scouring depth. Keywords: Artificial neural network, wavelet, local scouring, bridge pier, sensitivity analysis.

Actions (login required)

View Item