论文部分内容阅读
3层反向传播(BP)神经网络已用于预测火灾下平面管桁架钢的极限温度。网络模型的输入参数有直径比(β)、墙宽厚比(τ)、径厚比(γ)和荷载比,输出参数有极限温度。利用有限元软件ABAQUS建立神经网络模型。105组数据用于建立BP神经网络,15组数据用于测试和验证BP网络。建立BP网络的过程中,选用Levenberg-Marquardt反向传播算法。隐藏层选用tansig函数,输出层选用purelin函数。分析结果表明,使用BP网络模型预测的极限温度是准确有效的。
The 3-layer backpropagation (BP) neural network has been used to predict the ultimate temperature of flat tube truss steel under fire. The input parameters of the network model are diameter ratio (β), wall thickness ratio (τ), diameter to thickness ratio (γ) and load ratio. The output parameters have the limit temperature. Using the finite element software ABAQUS to build the neural network model. 105 sets of data for the establishment of BP neural network, 15 sets of data for testing and verification BP network. In the process of establishing BP network, Levenberg-Marquardt backpropagation algorithm was chosen. Hidden layer selection tansig function, output layer selection purelin function. The analysis results show that the limit temperature predicted by BP network model is accurate and effective.