基于范例推理的公路隧道拱顶变形时序支持向量机外延预测

Support vector machine extrapolation prediction of tunnel arch-top deformation time series based on case-based reasoning

针对公路隧道拱顶变形预测模型的普适性与外推预测的准确性,提出了基于人工智能推理的隧道工程属性(地理位置、监测位置、隧道高宽比、围岩级别和埋深)与拱顶变形时序曲线原子矩阵的相似范例检索方法,并在深入分析了获取的相似范例特征的基础上,进一步以LPG新核函数支持向量机建立先验知识的预测模型。应用该方法对通渝隧道工程K19+994断面拱顶下沉进行了预测与评估。结果表明,对于不同隧道间或同一隧道不同区段预判拱顶变形或收敛,基于范例推理能够获知良好的先验背景知识,且以此进行的支持向量机预测模型学习的回归内插(1～14步序)的平均相对误差为1.36%,而一次性外推预测15 d内的8个变形值(16～30步序)的平均相对精度为97.28%,证实了方法的可靠性。

An artificial intelligence reasoning and similar case retrieval methodology for tunnel projects characteristics, such as geographic region, monitor position, depth-width ratio, levels of surrounding rock mass and embedded depth of tunnel, and time series of arch-top deformation, is presented aimed at universality of prediction model and extrapolative accuracy. The similar case is searched and analyzed; and a forecast model is further built based on prior resource with LPG new kernel of support vector machine. The efficiencies of method are tested by predicting arch-top subsidence of a tunnel project example. Experimental results show that the approached method is suitable for the study of different tunnels or different sections of same tunnel. The average relative error of regression is 1.36% and the accuracy of extrapolation prediction is up to 97.28% within 8 steps proved case-based reasoning could achieve good prior resource and greatly help to enhance the ability of generalization of support vector machine. The better accuracy and reliability for extrapolative assessment of arch-top deformation can be utilized to server tunnel monitoring measurement.