针对大坝工作条件复杂,影响因素繁多,致使现有监控模型预报精度偏差过大问题,基于递阶对角神经网络能够逼近任意非线性函数的特点,使用串并联模型辨识器,采用动态BP学习算法,以水压、温度和时效因子为输入量,坝体位移为输出量,结合工程实例提出了大坝变形监测的递阶对角神经网络模型,并将该模型用于坝体变形数据的拟合分析及其预测预报。研究表明,该网络不仅收敛速度快,提高了算法的效率,而且对实测数据具有较好的拟合效果,提高了预报精度,在大坝安全预测分析中具有有效性和优越性。
In order to deal with dam monitoring data more effectively, a new deformation monitoring model has been proposed based on hierarchical diagonal neural network (HDHH) that can approximate any nonlinear function. The model takes water pressure, temperature and time factors as the input and dam displacement as the output. And then the series-parallel model identifier and dynamic BP learning algorithm play an important role in modeling. The dam deformation data fitting analysis and forecasting research show that the HDNN model is not only convergent quickly enough to improve the efficiency of the algo-rithm, but also has a good effect in fitting with monitoring data ,which improves forecast accuracy a lot. In a word, HDNN has great validity and superiority in the dam safety forecast analysis.