Please use this identifier to cite or link to this item: http://localhost:8081/xmlui/handle/123456789/8355
Title: APPLICATION OF ARTIFICIAL NEURAL NETWORKS FOR DETERMINATION OF WIND PRESSURE DISTRIBUTION IN BUILDINGS
Authors: Bitsuamlak, Girma Tsegaye
Keywords: CIVIL ENGINEERING;ARTIFICIAL NEURAL NETWORKS;WIND PRESSURE DISTRIBUTION;BUILDINGS
Issue Date: 1998
Abstract: Usually the evaluation of wind load on buildings is carried out using codes and standards, whose specifications are generally based on wind tunnel tests performed on selected building structures with common shapes. For example IS:875 (part-3) 1987 gives specification for rectangular, square, cylindrical, and uniform shapes. Wind pressure determination in buildings having shapes different from those specified on codes and standards requires wind tunnel studies. However wind tunnel studies can be expensive in terms of resources and time. Even wind tunnel facilities might not be available also. This dissertation suggests the use of Artificial Neural Network (ANN) for determination of wind pressure distribution in buildings using limited wind tunnel data. For this purpose, two ANN programs have been developed using C++. The first, based on back-propagation learning network (BPLN) and the other, based on cascade-correlation learning network (CCLN). Comparative study has been made using real data: for comparison between the two ANN algorithms and for paving a ground on application of ANN in Wind Pressure distribution determination in buildings. The back-propagation learns from training data at slow rate. In addition the optimal size of the back-propagation network can not be predetermined. Since BPLN is a fixed size training algorithm, the art of determining the size, depth and topology of the network is done most of the time by trial and error. On the other hand, CCLN determines its own size, depth and topology. It learns fast and retains the structure it has built. Moreover it is an incremental learning technique. Hence it is advantageous than BPLN. Wind Tunnel Studies on Pressure Distribution for Scope Twin Tower Office Building, from Wind Engineering Center, University of Roorkee, have been used in the present study. ANN models have been successfully trained using CCLN to predict pressure coefficients of the same.
URI: http://hdl.handle.net/123456789/8355
Other Identifiers: M.Tech
Research Supervisor/ Guide: Godbole, P. N.
metadata.dc.type: M.Tech Dessertation
Appears in Collections:MASTERS' THESES (Civil Engg)

Files in This Item:
File Description SizeFormat 
CED 248081.pdf5.67 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.