基于IPSO-MOC-RF算法的隧道智能支护方法研究

Research on intelligent tunnel support method based on IPSO-MOC-RF algorithm

面对隧道支护参数设计与实际地质条件匹配度较低的“通病”,智能支护理念呼声日趋增高,但实际研究与实施较少,建立智能支护决策方法对解决此问题具有重要意义。综合考量支护参数多样性与智能算法的适应性,通过改进随机森林算法与粒子群优化算法,提出一种适配多标签多输出数据类型的IPSO-MOC-RF智能决策算法。以勘察设计阶段、施工阶段及岩爆灾害为研究场景,建立支护参数分类指标体系;结合工程成功案例分别收集285,480和543份数据样本用于训练模型,MR值分别为0.886,0.917和0.897,汉明损失值为0.046~0.091;训练效果相比传统算法有明显提升,验证了该模型能有效提取地质指标与支护参数之间的复杂非线性映射特性。此外,通过Cesium平台集成BIM模型和智能决策模型建立隧道智能支护云平台,并应用于峨汉高速大峡谷隧道;采用统计对比分析、理论分析与现场连续监测的方法验证智能决策结果的有效性。结果表明:勘察设计阶段的7个样本中仅有1个结果失效;施工阶段的47个断面中仅有3个结果失效,选用的3个试验段全部成功,拱顶最大位移变形量为36.4 mm,边墙为25.8 mm;岩爆灾害的3个试验段均成功应用。模型应用效果显著,验证了该方法工程实际应用的可行性,可为隧道智能化建造的研究提供一种新思路。

The idea of intelligent tunnel support is an effective solution to the problem of low matching degree between tunnel support parameters and actual geological conditions.However,the research and implementation of tunnel intelligent support method are few,and the establishment and application of this method is in great demand.In this paper,by improving random forest algorithm and particle swarm optimization algorithm,an IPSO-MOC-RF intelligent decision algorithm adapted to multi-label and multi-output data types was proposed to analyze the complex and diverse support parameters of tunnels.The tunnel support parameter system was established according to three research scenarios of tunnel survey and design stage,construction stage and rock burst disaster,and 285,480 and 543 data samples were collected respectively for training models.According to the training results,the MR values were 0.886,0.917,0.897,and Hamming Loss values were 0.046-0.091,respectively,which verified that the model could effectively extract the complex nonlinear mapping characteristics between geological indicators and tunnel support parameters.In addition,a tunnel intelligent support platform was established by integrating BIM model and intelligent decision model with Cesium cloud platform,and applied to the Grand Canyon tunnel of the Ehan expressway in China.The validity of the intelligent tunnel support decision is verified by statistical analysis,theoretical analysis and field continuous monitoring.The results show that only 1 of the 7 samples in the survey and design stage failed;only 3 of the 47 samples in the construction stage failed,and the 3 selected test sections were all successful(the maximum displacement of the tunnel vault and side wall were 36.4 mm and 25.8 mm,respectively);the three test sections of rockburst disasters were successful.The application effect of intelligent tunnel support model is remarkable,which verifies the feasibility of the practical application of this method,and provides a new idea for the research of intelligent tunnel construction.