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

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。

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

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 ef...