基于神经网络的混凝土大坝弹性参数识别方法

Parameter identification of concrete dam with neural networks

基于人工神经网络方法 ,根据云峰大坝坝顶水平位移观测资料识别大坝混凝土和岩石基础的弹性模量 .采用修正的BP学习算法 ,并通过对迭代步长的优化计算及对观测数据的归一化处理 ,提高了参数识别的速度和精度 .云峰大坝的工程实际应用表明 ,用神经网络方法识别材料参数具有识别精度高和收敛速度快等特性 ,拟合误差小于 0 .15mm .

Based upon the artificial neural networks ,the elastic moduli of the concrete dam and the rock foundation are identified according to the horizontal displacements of the concrete dam in the Yunfeng projects. The precision of the estimated parameters and convergence are enhanced by the optimum of the iteration step size and handling in advance for the observed data. The practical application to the concrete dam in the Yunfeng projects indicates that the parameter identification algorithm proposed by this paper possesses higher identification accuracy, faster convergence, and the fitting error is less than 0.15 mm.