BP神经网络融合预报在深基坑监测中的应用

Application of BP Neural Network Fusion Forecast for Deep Excavation Deformation Monitoring

针对多传感器采集的多种变形信息,提出了利用BP人工神经网络的方法进行数据融合处理,以达到对变形体变形状况更加准确预报的目的。通过BP神经网络训练精确得到各种观测值权重,避免了人为主观因素的影响。利用该方法对多传感器数据进行横向预报可以实现对暂时损坏的传感器数据准确预测,保证数据的连续性。也可以采用该方法进行多传感器数据纵向融合预报。在广州某小区基坑综合监测项目中,预报结果与实际变形情况相吻合,取得了良好的效果。

This paper proposed a method based on Error Back Propagation Neural Network to syncretize the multi-sensor deformation data to precise forecast the deformation, and to obtain the observed value＇ s precise weight through the Error Back Propagation Neural Network＂s train to avoid subjective influence. This method could precisely find the broken sensors by across forecast of multi-sensor＇ s data, to make sure the data continuity. It could also use for longitudinal forecast. In the deep excavation monitoring engineering in Guangzhou, the ordinal forecast result of multi-sensor＇s data was consistent with the real deformation, achieved good effect.