基坑变形灰色人工神经网络预测模型及其应用

Grey Artificial Neural Network Prediction Model for Deformations of Foundation Pits and Its Application

针对基坑变形预测中信息的灰色性和数据的非线性性,提出用灰色神经网络预测基坑变形的新方法。用一桩锚联合支护体系实例进行了预测研究,得到支护体系的不同预测模型的组合预测值。研究结果表明:灰色神经网络预测误差比GM(1,1)预测模型小;与BP预测模型相比,前期误差大,后期误差小。在基坑变形监测中,为了更准确地预测基坑变形,可以采用灰色神经网络预测与BP预测相结合的方法进行预测。

Considering the greyness of information and nonlinearity of datain deformation forecast for afoundation pit,anew method,the grey artificial neural network（GANN）,for predicting the deformation of afoundation pit was proposed.Taking the foundation pit supported by acombined system of piles and anchors as an example,the deformations were predicted by different forecast models of the GANN.The results show that the errors of grey artificial neural network forecast model are smaller than that of GM（1,1）model,and bigger in early days and smaller in later period than that of BP model.In order to predict the deformation of foundation pit more accurately,the grey artificial neural network forecasting model should be used combining with the BP model.