深部煤层非连续区地应力场反演方法  

Inversion method of stress fields in the discontinuous zone of deep coal seam

作　　者：周家兴 王金安[4,5] 李飞 ZHOU Jiaxing;WANG Jin-an;LI Fei(Department of Hydraulic Engineering,Tsinghua University,Beijing 100084,China;State Key Laboratory of Hydroscience and Engineering,Tsinghua University,Beijing 100084,China;Key Laboratory of Hydrosphere Sciences of the Ministry of Water Resources,Tsinghua University,Beijing 100084,China;School of Civil and Resource Engineering,University of Science and Technology Beijing,Beijing 100083,China;Key Laboratory of Ministry for Efficient Mining and Safety of Metal Mines,University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]清华大学水利水电工程系,北京100084 [2]清华大学水圈科学与水利工程全国重点实验室,北京100084 [3]清华大学水利部水圈科学重点实验室,北京100084 [4]北京科技大学土木与资源工程学院,北京100083 [5]北京科技大学金属矿山高效开采与安全教育部重点实验室,北京100083

出　　处：《清华大学学报（自然科学版）》2024年第12期2166-2176,共11页Journal of Tsinghua University(Science and Technology)

基　　金：国家重点研发计划(2017YFC1503104);中国石油大学油气资源与工程全国重点实验室(PRP/open-2213);北京科技大学青年教师学科交叉研究项目(FRF-IDRY-21-025)。

摘　　要：随着煤矿开采深度的增加,断层等非连续构造活化会对煤层安全高效开采带来巨大隐患,掌握精确的地应力场分布对于开采工程设计与施工、防灾减灾具有重要的指导意义。针对深部地应力场的非线性及非连续性特征,推导了正断层、逆断层和走滑断层区域地应力侧压力系数的稳定性判别式,并基于长短时记忆神经网络算法,提出了深部煤层非连续区的地应力场反演方法。该方法将不同时期的地应力资料作为时间序列进行优化学习,有效解决了深部实测地应力数据的非线性、离散性、多噪声等问题,确保了优良地应力数据信息长期记忆,不良地应力数据信息及时遗忘。以上海庙矿区鹰骏二号井田主副井区域为例,考虑了不同尺度断层的地应力场分布特征,反演得到详尽的地应力场分布。反演与实测地应力场的相关系数为0.945,平均误差为12.897%,地应力差值的标准差为2.000。因此,该方法可为深部煤层非连续区地应力场反演提供新的思路。[Objective] As the depth of coal mining increases,the activation of discontinuous structures,such as faults,poses a significant risk to the safe and efficient mining of coal seams.Therefore,acquiring precise knowledge of the distribution of in-situ stress is paramount for the design,construction,and disaster prevention of mining engineering.[Methods] This study proposes an inversion method for in-situ fields applicable to discontinuous zones of deep coal seams.(1) Given the discontinuity characteristics of the deep in-situ stress field,stability discriminants for normal faults and stability discriminant equations for positive faults,reverse faults,and strike-slip fault zones are derived based on the lateral pressure coefficients of in-situ stress.(2) A long short-term memory neural network algorithm is adopted to optimize the learning of the in-situ stress field data formed in different periods sequentially to effectively solve the nonlinearity,discreteness,and multi-noise problems of the measured deep in-situ stress data and to ensure that the excellent in-situ stress data information is remembered for a long time and that the inferior in-situ stress data information is forgotten in time.[Results] This study considers the main and auxiliary well areas of Yingjun's second mining area in Shanghai Miao as the research background and establishes an algorithm model for long short-term memory neural networks.Given the distribution characteristics of the in-situ stress field in fault areas at different scales,an inversion calculation of the in-situ stress field in discontinuous areas of deep coal seams was conducted.[Conclusions] The correlation coefficient between inverted and measured stress fields was 0.945,with an average error of 12.897%.The standard deviation of the stress difference is 2.000.The amount and direction of the regional stress field of the fault will also change.Compared with the regional stress field,the in-situ stress field in the DF15 and SF15 large-scale fault zones is approximately 5 MPa lower,a

关 键 词：深部煤层 非连续 地应力场 反演方法 

分 类 号：TU45[建筑科学—岩土工程]