径向基神经网络在沉降预测中的应用

Application of radial basis function neural networks in settlement predictive model

针对大型建筑物的沉降监测存在较多的不等时间间隔沉降监测数据,现有模型需要对此类数据进行等间隔处理后建立沉降预测模型的现状,该文采用无需对监测数据进行等间隔预处理的径向基神经网络对沉降监测数据建立沉降预测。通过对西安某大厦基坑开挖对地表和周围建筑物影响的沉降监测数据进行实例分析,并与非等间隔灰色GM(1,1)预测模型进行对比,利用模型评价指标评价预测模型精度。结果表明:采用径向基神经网络建立预测模型处理过程简便,其预测精度优于非等间隔灰色GM(1,1)预测模型。

There are a range of time intervals of subsidence monitoring data in the practical settlement monitoring of large buildings. The application intervals need to be processed when such data are used, however, they don＇t need to be preprocessed when establishing the simulation model by using radial basis function （RBF）neural network to predict the subsidence monitoring data. This paper analyzed the monito- ring data of a building＇s foundation pit excavation on the surface and the surrounding buildings settlement in Xi＇an, compared with the commonly used non-interval grey GM （1, 1） prediction model, and evalua- ted forecast model accuracy using MAPE and PVR, it was proved that RBF neural network was more con- venient and more accurate than the interval grey GM （1, 1） prediction model.