Please use this identifier to cite or link to this item:
標題: Application+of+Neural+Networks+for+Modeling+Shear+Strength+of+Reinforced+Concrete+Beams
作者: 湯兆緯
關鍵字: 鋼筋混凝土樑;剪力強度;類神經網路
出版社: 國立中興大學工學院;Airiti Press Inc.
Project: 興大工程學刊, Volume 13, Issue 3, Page(s) 153-170.
Reinforced concrete is a nonhomogeneous, nonisotropic, and highly nonlinear material, so the real distribution of shear stresses over its cross section is a complicated problem that makes uneasy in mathematical modeling. Based on the artificial neural networks technology, this paper presents a nontraditional approach to the prediction of the ultimate shear strength of RC beams with web reinforcement. A standard back-propagation neural network (BPNN) and a multilayer-functional-link neural network (MFLNN) are used for training and testing the experimental data. A comparison study between the neural network model and five parametric models is also carried out. It was found that both the BPNN and the MFLNN are able to generalize the functional relationship between the independent variables and the measured dependent variables, and the MFLNN has better accuracy and efficiency than the BPNN. Moreover, compared with parametric models, the neural network approach provides better results. The results show that neural networks have strong potential as a feasible tool for predicting the ultimate shear strength of RC beams with web reinforcement.

鋼筋混凝土是一種非均質、非等向性與非線性之材料,在承受荷重狀況下,其斷面內的剪應力分佈相當複雜,數理模式不易建立。基類神經網路(artificial neural networks)技術,本文提出異於傳統的研究方法,以預測含腹筋鋼筋混凝土樑的極限剪力強度。文中採用標準的倒傳遞類神經網路(back-propagation neural network,簡稱BPNN)與變良的倒傳遞類神經網路-多層函 數連結網路(multilayer-functional-link neural network,簡稱MFLNN),以訓練與測試實際範例,並將建構完成的類神經網路評估模式之預測值與現 鋼筋混凝土樑剪力分析模式之預測值作比較。研究結果顯示,應用類神經網路可有效預測含腹筋鋼筋混凝土的剪力強度,且MFLNN之準確性與效率均優於BPNN;此外,類神經網路預測值的準確性也比既有經驗公式來得精確。由有結果可知,類神經網路具有深厚潛力以發展成為預測含腹筋混土樑剪力強度的一種可行工具。
ISSN: 1017-4397
Appears in Collections:第13卷 第3期

Files in This Item:
File SizeFormat Existing users please Login
142315-1.pdf1.05 MBAdobe PDFThis file is only available in the university internal network    Request a copy
Show full item record

Google ScholarTM


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