改进后灰色神经网络及短期基坑沉降预测研究

Improved grey neural network and research on prediction of short-term foundation pit settlement

在短期基坑沉降监测中,由于数据量少且呈非线性变化,沉降模型很难准确建立。灰色GM(1,1)对数据少、趋势性强、波动小的数据有较高的预测精度,但不能模拟复杂的非线性函数;BP神经网络可以对非线性数据进行学习训练,具有自学习、自适应能力;通过将GM(1,1)与BP神经网络组合,并优化网络部分的学习率、权值和阈值等,建立一种改进的灰色神经网络模型,该模型具有对非线性数据自学习、自适应能力和预测精度更高等优点。通过某基坑沉降监测分析,验证改进的灰色神经网络模型预测精度更高,适合短期建模,具有很好的实用性。

For foundation pit settlement in the short term ,the settlement prediction model is difficult to be established because of the data is less and nonlinear .GM (1 ,1) has a high predictive accuracy to cope with less ,clear trend and little fluctuation data .But it’s of low precision for the complex nonlinear function .The BP neural network can learn well the nonlinear data ,which has good self‐learning and adaptive ability .An improved grey neural network is established by combining GM (1 ,1 ) with BP neural network ,and optimizing the network learning rates ,weights and thresholds .The improve model has well self‐learning adaptive ability and higher prediction accuracy .In one actual foundation pit project ,the fact proves the improved grey neural model is higher precision ,suitable for short‐term modeling ,w hich is of very practicality .