BP网络在大坝变形空间多测点监测模型中的应用

Application of BP Network to Multiple-spot Model of Dam Deformation Monitoring

利用BP网络的较强函数非线性映射能力,对大坝多个测点的变形监测数据进行了仿真及预测,所建立的空间多测点BP网络模型以水位、温度、时间、测点坐标作为输入参数,以大坝变形值作为网络输出。结果表明,通过有代表性的学习样本对网络的训练,模型的拟合及预报结果能够满足精度要求,同时,与传统的统计建模方法相比,大坝变形多测点BP网络模型具有实施容易、运行速度快、误差修正方便、操作简单、运行灵活等优点。

This paper carries out simulation and forecast of deformation monitoring data of multiple-spot of dams by utilizing strong function non-linear mapping capability of BP network and creates the multiple-spot BP space model with input parameters of coordinates of water level, temperature, time and measuring spots and output of dam deformation values. The results reveal that the simulated and forecasted results can meet monitoring accuracy requirements through typical sample cases of network training. Meanwhile, multiple-spot BP network model of dam deformation demonstrates the advantages of easy model execution, fast running speed, convenient error revision, simple and flexible operation in comparison with traditional statistical modeling practice.