车 轶¹，  尤  杰²，  徐东坡³，  仲伟秋¹
    （1 大连理工大学建设工程学部，大连116024；2 中兴通讯股份有限公司CAF规划系统部，深圳518055；3 中国人民解放军73670部队，南京210009）

[摘要] 在分析5种具有不同输入变量的神经网络模型的基础上，建立了钢筋混凝土无腹筋梁抗剪强度的优化人工神经网络模型。该模型具有4个输入变量（混凝土抗拉强度、剪跨比、纵筋配筋率和截面有效高度）和一个输出变量（抗剪强度）。通过对数据的放缩处理，提高了网络训练效率。此外还对我国GB50010-2002规范、ACI318 -08规范、Eurocode2、日本JSCE规范和加拿大CSA A23. 3-04规范的无腹筋梁抗剪计算公式进行了对比研究。研究表明，神经网络模型具有较高计算精度，能够很好地预测无腹筋梁的抗剪强度。在各国规范公式中，CSA A23. 3-04．规范的计算结果与试验结果吻合很好，我国GB50010-2002规范、ACI318-08规范和Eurocode2公式计算结果的离散性较大。
[关键词] 人工神经网络；抗剪强度；混凝土无腹筋梁
中图分类号：TU375.1    文献栎识码：A  文章编号：1002-848X(2011) S2-0223-06

    Artificial neural network model for sheer strength of reinforced concrete beams without web reinforcement
    Che Yi¹,  You Jie²,  Xu Dongpo³,  Zhong Weiqiu¹
    (1  Faculty of Infrastructure Engineering,  Dalian Uruversity of Technology,  Dalian 116024,  China; 2 ZTE Corporation CAF Planning & System Department, Shenzhen 518055, China; 3 People's Liberation Army 73670 Troops,  Nanjing 210009,  China)

Abstract: An optinuzed artificial neural network( ANN)  model for predicting shear strength of reinforced concrete beams without web reinforcement was presented on the basis of study on 5 ANN models with different input variables. The artificial neural network model has four input variables,i.e.concrete tensile strength, shear-to-span ratio, longitudinal reinforcement and effective depth, and an output variable, i.e. shear strength. By scaling the data processing, the efficiency of network training was improved.  Predictions of shear design equations in GB50010-2002 code, ACI318-08 code, Eurocode2, JSCE code and CSA A23.3-04 code were also compared with test data. It is found that the proposed artificial neural network model predicts shear strength of beams without web reinforcement with a very good accuracy. Among the code equations,  CSA A23. 3-04method fits tests data well, wlule equations in GB50010-2002 code,  ACB18-08 code and Eurocode2 present large scatters of results.
Keywords: artificial neural network; shear strength; concrete beams without web reinforcement