摘要
为充分发挥监控量测在隧道施工过程中的作用及价值,将那朗隧道作为工程实例背景,以支持向量机为理论基础,并利用粒子群算法、混沌理论等优化其模型参数,进而构建出混沌优化PSO-SVM模型,以实现隧道变形的准确预测;利用重标极差法判断隧道变形的发展趋势,以佐证前述预测效果的准确性。实例研究表明:通过试算法和粒子群算法能有效优化支持向量机的模型参数,且混沌理论能有效弱化预测结果的残差序列,所得预测结果的相对误差均值均小于2%,验证了本研究预测模型的有效性;同时,重标极差分析得出隧道变形虽会持续增加,但增加速率趋于减小,所得结果与预测结果相符,验证了前者分析结果的准确性。研究发现,为隧道变形预测提供了一种新的思路。
In order to give full play to the role and value of monitoring and measurement in the tunnel construction process,the Nalang Tunnel was used as the background of the engineering example.The support vector machine was used as the theoretical basis,and its model parameters were optimized by using particle swarm optimization and chaos theory.A chaos optimized PSO-SVM model was constructed to achieve accurate prediction of tunnel deformation.The re-calibration range method was used to judge the development trend of tunnel deformation to prove the accuracy of the aforementioned prediction effect.The case study showed that the model parameters of the support vector machine could be effectively optimized by the trial algorithm and the particle swarm algorithm,and the chaos theory could effectively weaken the residual sequence of the prediction results.The average relative errors of the prediction results were less than 2%,which validated the research.The validity of the prediction model;at the same time,the re-standard range analysis showed that the tunnel deformation will continue to increase,but the increase rate tended to decrease.The obtained results were consistent with the prediction results,which verified the accuracy of the former analysis results.The study found that it provided a new idea for tunnel deformation prediction and was worth further promotion and application.
作者
付俊生
FU Junsheng(No.3 Engineering Co.,Ltd.,CCCC First Harbor Engineering Co.,Ltd.,Dalian 116083,Liaoning,China)
出处
《隧道与地下工程灾害防治》
2019年第4期103-108,共6页
Hazard Control in Tunnelling and Underground Engineering
关键词
隧道
变形预测
支持向量机
粒子群算法
重标极差分析
tunnel
deformation prediction
support vector machine
particle swarm optimization
rescaling range analysis
