基于神经网络的矩形钢管高强混凝土计算方法

摘要

Structures of concrete-filled rectangular steel tubular columns (CFT) have been widely used throughout the world.Calculation methods have existed to calculate the strength of CFT in many standards,but calculation results vary with different standards.Not all standards are suitable to design CFT as the strength of concrete used in the project is becoming higher and higher.This paper collects data of axial loaded CFT experiments reported in recent papers,and trains 2 models (ANN-1 and ANN2) with different input parameters and bearing capacity as output parameter using the data.Accurate and stable ANN models are verified and tested.Finally,parameter analysis is conducted in ANNs,and its value is compared with the normalized value.The results demonstrate ANN models are available for the prediction of bearing capacity of high strength concrete-filled rectangular steel tubular columns.%矩形钢管混凝土结构在国内外已经广泛应用,许多规范中也规定了其计算方法,但是各个规范中矩形钢管混凝土的计算结果并不一致,并且随着工程中混凝土强度越来越高,并不是所有的规范都适用于矩形钢管高强混凝土计算.基于神经网络方法,搜集近年来国内外矩形钢管高强混凝土轴压短柱试验数据,并用这些数据训练了两个以不同参数为输入、以承载力为输出的神经网络模型ANN-1和ANN-2,经过调试和验证,得到了准确和稳定的模型.最后对矩形钢管高强混凝土进行了参数化分析,并与规范计算值进行对比,评价了规范中的计算方法,并进一步证明模型可以用于矩形钢管高强混凝土承载力的预测.