支持向量机在地铁隧道地表沉降预测中的应用

Application of Support Vector Machine for Subsidence Prediction of Subway Tunnels

针对北京地铁区间浅埋暗挖隧道,分析了施工导致的地表沉降的影响因素,提出了影响地表变形的主要因素,建立了支持向量机的统计回归预测模型,并用该模型预测的地表沉降,将SVM预测值和BP神经网络预测值与现场量测值进行了对照。结果表明,支持向量机比BP神经网络有较高的预测精度,并且具有小样本、高维数及非线性等优点。

Some influence factors of surface subsidence due to shallow excavation construction in Beijing subway tunnel were analyzed, and the numerical prediction model based on SVMs were founded. The model is used to predict the surface displacements of subway tunnel and compared with the BP neural networks predictions and with measure datum in site. The result indicates, that the accuracy of predictions by SVMs is better then the predictions of BP neural networks, and with some merits of small samples, multi-dimensions and non-linear.