基于NNs模型的轻骨料混凝土梁抗剪分析  

Shear Capacity of Reinforced Lightweight Concrete Beams Using Neural Networks

作　　者：张玉[1] 岳志豪 刘伯权[1] 

机构地区：[1]长安大学,西安710061

出　　处：《硅酸盐通报》2016年第6期1780-1784,共5页Bulletin of the Chinese Ceramic Society

基　　金：国家自然科学基金项目(51578072;51578077);陕西省科技统筹创新工程计划项目(2015KTZDSF03-04);中央高校基本业务费(310828163410;310828161006)

摘　　要：混凝土构件的抗剪问题始终未形成统一定论。优化多层前馈神经网络(NNs)模型使用反向传播算法及提前终止技术,能够合理考虑各层神经元几何与材料特性。基于神经网络建立了轻骨料混凝土梁的受剪承载力计算模型,并搜集国内外82组轻骨料混凝土梁受剪试验结果作为样本数据,分为训练组、验证组及测试组,通过与试验值对比分析验证了计算模型的合理性和准确性。研究表明:训练组、验证组及测试组的试验值与NNs模型计算值比值的平均值分别为0.953、1.064和1.124,方差为0.147、0.034和0.091,NNs模型的计算结果能很好的对轻骨料钢筋混凝土梁的抗剪承载力进行预测,并能充分考虑各影响因素的显著性。Shear resistance of concrete members has been discussed a lot.The model of multilayered optimized feed forward neural network（NNs） using back-propagation algorithm and the early termination technology can consider geometric and material properties of neurons of each layer reasonably.Based on neural network,the shear capacity calculation model of lightweight aggregate concrete beams is established.In this paper,the sample data of 82 group test results of lightweight aggregate concrete beams under shearing collected from home and aboard are divided into training groups,the validation group and the test group.And the test results are compared and analyzed with test results to verify the rationality and accuracy of the calculation model.The study shows that The averages ratio of the test values of training groups,the validation group,the test group and the value of NNs models are respectively 0.953、1.064 and 1.124,and the variance are respectively 0.147,0.034 and 0.091.The calculation results of NNs model can forecast the shear capacity of lightweight aggregate concrete beams well,and can fully consider the significance of each factor.

关 键 词：轻骨料混凝土 梁 受剪承载力 神经网络 

分 类 号：TU528[建筑科学—建筑技术科学]