摘要
围岩变形预测是隧道安全评价及其指导后期施工的重要依据,为提高变形预测精度,结合工程实践,提出了PSO-SVM-BP预测模型的思路。首先,利用三次样条插值及二次平滑法对变形数据进行预处理,为后期变形预测奠定基础;其次,利用粒子群算法对支持向量机进行参数优化,建立PSO-SVM模型,并对围岩变形进行初步预测;最后,利用BP神经网络进行误差修正,达到综合预测的目的,并利用工程实例进行检验,以验证预测模型的有效性。结果表明:初步预测结果的相对误差均小于5%,而误差修正后的预测精度被提高到0.97%,预测精度较高,验证了预测模型的有效性,可为类似研究提供参考。
The prediction of surrounding rock deformation is an important basis for the safety evaluation of the tunnel and the construction of the later stage. In order to improve the precision of the deformation prediction, by combining with the engineer-ing practice, the idea of PSO-SVM-BP prediction model is put forward. First of all, the deformation data are pre processed by three spline interpolation and smoothing method for two times, laying the foundation for the late deformation prediction; secondly, to optimize the parameters of support vector machine based on particle swarm algorithm, then PSO-SVM model is established, and the surrounding rock deformation is predicted preliminarily; at last, a BP neural network for error correction is used to achieve comprehensive forecasting purposes, and engineering examples are used for the test to verify the effectiveness of the prediction model. The results show that the relative error of preliminarily prediction results is all less than 5 % , and the prediction accuracy after error correction increases to 0. 97%, showing higher prediction accuracy, which proves the validity of the forecast model. The prediction model is feasible, and can provide a reference for similar research.
作者
任庆国
苗兰弟
REN Qingguo MIAO Landi(Shaanxi Railway Institute, Weinan, Shaanxi 714000, China)
出处
《河北工业科技》
CAS
2017年第3期194-201,共8页
Hebei Journal of Industrial Science and Technology
关键词
隧道工程
粒子群算法
支持向量机
BP神经网络
动态预测
tunnel engineering
particle swarm algorithm
support vector machine
BP neural network
dynamic prediction