高速公路连拱隧道施工变形预测的GA-SVR智能模型研究

基于连拱隧道支护结构繁多，施工工序复杂，同时由于施工影响，现场监测数据较少且数据误差较大，造成传统方法很难用于连拱隧道的施工变形预测。针对于此，支持向量回归（SVR）算法可以任意精度逼近任意函数，与神经网络相比具有小样本、全局优化和泛化性能好的优点。本文结合铜黄高速公路富溪连拱隧道的施工变形监测，采用遗传算法来优化支持向量回归算法的模型参数，形成GA-SVR算法，建立起了连拱隧道变形预测的GA-SVR智能模型。采用此模型对富溪隧道后继开挖的监测时间点进行变形预测，并与GA-BP模型对比可以看出本文所建立的GA-SVR智能模型具有极高的预测精度，完全可用于连拱隧道施工期的变形预测，也为类似工程提供了借鉴。

Study on GA-SVR Intelligent Model for Deformation Prediction of Expressway Multi-arch Tunnel Construction

Deformation prediction during multi-arch tunnel construction period is very difficult because of its various structure and complex operation procedure.At same time,the traditional methods cannot be used to forecast the displacement of multi-arch tunnel due to lack of monitoring data usually.The support vector regression(SVR) algorithm,which embodies the characteristics of applicability for small sample,global optimization and better generalization capability than artificial neural network,can approach any f...