BP和Elman神经网络在砂土液化预测中的研究

摘要

Based on the condition that the influencing factors of sand liquefaction have a non-linear relationship,while the neural network model can simulate any nonlinear function and is suitable for dynamic system recognition,in this paper,a three-layer BP neural network and Elman network with 4 input layer neuron,2 hidden layer neuron and loutput layer neuron were established respectively.What's more,by the way of practical examples simulations,Elamn neural networks haa a faster convergence speed,a higher accuracy and a better effect than Elamn neural networks.%基于砂土液化的影响因素具有非线性关系,而神经网络模型能够逼近任意非线性函数和适合于动态系统辨识的特性,分别建立输入层为4,隐含层神经元为2,输出层为1的三层BP神经网络和Elman网络,并且通过matlab软件运算,实例比较得出Elman模型比BP模型收敛速度快、精度高,在砂土液化的预测中效果更好.