诱发型岩爆的内外因参量贡献机制及预警模型

Prediction of rockbursts is an urgent yet challenging research topic in deep rock engineering. Given the suddenness and the complexity of rockbursts, it is difficult to propose an accurate and universal criteria for rockburst prediction. The current rockburst prediction level cannot meet the engineering requirement of the excavation in deep hard rock with such a high efficiency. First, Induced rockbursts factors caused by blasting disturbance, shape and size of chambers, rock mass and geostress characteristics will be selected in this work based on the existing engineering structure of deep projects. The frequency spectrum characteristics and the stress wave attenuation laws of the disturbed sources of blasting will be achieved by Hilbert-Huang Transform and mathematical statistics methods with via the on-site monitoring data. Second, the quantitative characterization of equivalent loads of blasting disturbance is obtained. The numerical simulation model of rockburst risk is acquired using the Monte Carlo, the Response Surface and the Finite Difference method. The database of rockburst information is also established. Then, internal and external parameters in the process of rockburst contribution mechanism are revealed using quantitative association rules of data mining, then the rockburst multi-factor quantitative criterion and prediction model of induced rockbursts are also established using ensemble learning algorithms. Finally, non-parametric statistical tests and verification tests of the proposed rockburst prediction model are conducted. The results of this research can promote the improvement of the rockburst criterion and prediction method. Also, it can provide theoretical guidance for the control of the deep rockburst disaster.

岩爆预测是深地工程亟待解决且具有挑战性的前沿课题，岩爆诱发因素的复杂性和岩爆显现的突发性，很难提出准确通用的岩爆判别准则，当前岩爆预测水平还远不能满足深部硬岩安全高效开挖的工程要求。本项目立足深地工程既有工程结构，综合考虑诱发型岩爆灾害致因—爆破扰动、硐室形状和尺寸、岩体质量和地应力特征，通过现场监测扰动波形并结合希尔伯特-黄变换和数理统计方法，分别获取表征爆源的频谱特征和应力波衰减规律；进而推导爆破扰动等效荷载的定量表征方式，并结合蒙特卡罗、响应面和有限差分等方法，建立诱发型岩爆风险度量的数值仿真模型，获取岩爆信息大数据；利用量化关联规则的数据挖掘技术和集成学习算法揭示内外因参量在岩爆过程中的贡献机制，继而建立诱发型岩爆多因素定量综合判据与风险预测模型，并进行非参数统计检验及工程现场扰动诱发岩爆验证实验。研究成果可促进岩爆判据与预测方法的日臻完善，并为深部岩爆灾害防控提供理论指导。

岩爆预测是深地工程亟待解决且具有挑战性的前沿课题，岩爆诱发因素的复杂性和岩爆显现的突发性，很难提出准确通用的岩爆判别准则，当前岩爆预测水平还远不能满足深部硬岩安全高效开挖的工程要求。本项目立足深地工程既有工程结构，通过文献调研、理论分析、数值模拟和数据挖掘等方法，系统实录了近30年岩爆信息数据库，推导了爆破诱发岩爆数值模拟的解析解；综合考虑诱发型岩爆灾害致因—爆破扰动、硐室形状和尺寸、岩体质量和地应力特征，数值模拟合理准确地探究了内外参量对诱发型岩爆的贡献机制，利用量化关联规则的数据挖掘技术定量揭示了各指标在岩爆过程中的参与机制，发现应力水平、工程尺度和爆源特征为岩爆主要贡献参量，当应力水平超过某一范围时，对岩爆风险值的影响一直增加，而此时工程尺度/爆源特征的值越大，则对岩爆风险值的预测呈现出更大影响；在修正了部分传统岩爆判据阈值基础上，提出了多因素诱发型岩爆新判据RBi，建立了多指标打分的岩爆烈度综合评价系统和基于Kriging空间插值的岩爆冲击倾向性分类方法；在数据挖掘阶段采用SVM，RF，LR，XGB和LGBM算法作为基学习器建立一系列岩爆预测模型；通过启发式算法和堆栈式学习方式建立了性能稳健的岩爆集成学习预测模型，对诱发型岩爆动力灾害进行预测预报，预测精度高达到90%以上。研究成果促进了岩爆判据与预测方法的日臻完善，为深地工程岩爆灾害准确评估、预测与防控提供了理论依据与技术支撑。