基于时间反转的煤矿微地震事件全波精确定位方法研究

Mine dynamic disasters pose a serious threat to the safety of mine industry. The design of microseismic monitoring systems for precisely locating fracture sources has important theoretical significance and application value to carry out accurate prediction work for coal-rock disasters. Time reversal based localization method exhibits high resolution focus and localization performance in acoustic field, which provides a new solving way to accurately locate microseismic events in mining. In order to solve the problem of coal mine microseismic events precise positioning is the basic starting point, combining with theoretical analysis, numerical simulation and experiments, the project will carry out the following research contents. After analysis of focal mechanism and its temporal evolution rule of coal-rock burst and propagation law of microseismic wave in coal-rock medium, the dynamic relationship between coal-rock failure process and the micro seismic response is established. Time reversal focusing performance is analyzed for microseismic wave in coal-rock medium based on the relationship, and then the time reversal focusing compensation model for microseismic wave in coal-rock medium is established. The purpose of the compensation model is to improve the resolution and accuracy of the anisotropy of the coal and rock medium micro wave time reversal focusing. A best compensation sensor array layout model is built with the interpolated array transform methods as guidance for an equivalent optimal observation system to reduce the requirements for observation array. The formulas for calculating location by time reversal method using full wavefields are derived. The algorithm are tested and optimized by numerical calculation and field trials to realize the automatic precise localization for coal-rock fracture events. The research can provide technical support for safety mining.

煤矿动力灾害严重威胁了矿山生产安全，开发高精度微地震定位方法，精确预测煤矿动力灾害的岩体微破裂前兆事件，对采矿安全工程具有重要的理论意义和实际价值。时间反转法在声学领域表现的高分辨聚焦定位性能，为解决微地震事件的精确定位提供了新的思路。本项目拟结合理论分析、实验研究、数值模拟等方法，以解决煤矿微地震事件精确定位问题为基本出发点，开展以下研究：分析煤岩体破裂诱发微地震的震源参数及其时空演变规律、构建传播方程并分析传播特征，揭示煤岩体破裂与微地震响应的动态关系；分析震源及介质物性参数对时间反转聚焦性能的影响，揭示煤岩体破裂诱发微地震波的时间反转聚焦性能，提出时间反转聚焦补偿模型，提高各向异性煤岩介质中微地震波时间反转聚焦的分辨率和精度；基于内插阵列变换等方法，提出最优观测系统补偿模型，降低定位方法对观测布阵的要求；推导全波精确定位公式，实现煤岩破裂源的精确定位，为矿山安全开采提供技术保障。

项目采用美国GCTS (Geotechnical Consulting and Testing Systems)公司的RTR系列的高温高压多功能岩石三轴测试系统的声发射检测技术，对煤样进行了声发射实验，通过对煤样重复施加应力的过程，记录了煤样在应力作用下，从一开始发生形变到煤样破裂这一完整过程，此过程由四个阶段组成：压密、弹性、屈服和破坏。获得声发射信号在应力-时间关系下的特征变化情况，描述了煤样受到不同压力和温度下声发射的特征。由于声发射事件个数的有效统计对于整个声发射试验有着至关重要的影响，为了提高声发射事件识别的智能性和准确率，引入了深度学习方法，利用7层LSTM相互堆叠构建了基于深度循环神经网络的声发射/微震事件自动识别网络。模型训练完成后，把进过声发射测试得到的波形数据，输入到训练好的模型中，得到相对应的波达特征，进而完成对煤岩体破裂全过程中声发射事件的特征分析和统计。该模型在多事件波达拾取上性能良好，而且对于一些人眼难以辨别的事件，模型提取的特征通过面分裂和面宽度的形式来表现事件的个数和相互叠加的情况。该模型识别事件的最高分辨率为5个采样点（实验采用的采样间隔为1μs，故分辨率高达5μs），不仅可以用于声发射/微震事件识别，还可以用于微震事件震相识别及到时拾取，有很好的应用价值和前景。利用双相介质参数对煤岩介质进行刻画，推导得到了该介质中的波动方程，结合声发射事件识别结果拟合得到了微震事件的震源函数，利用时域有限差分法，求解得到了煤岩破裂过程下的微地震响应。推导得到了时间反转定位法及其在不同观测系统参数下的聚焦性能，在理论上给出了以下结论：在固定检波器数量的情况下，阵列中距离最远的两个检波器与震源三者之间的夹角对聚焦效果影响最大，在一定条件下，聚焦效果与夹角的正弦函值成比例，为实现高精度回传波场聚焦提供合理依据。该结论通过数值实验得到了验证，为最优观测系统的设计提供了依据，并结合微震数据滤波处理和空间插值方法实现了最终的微震定位模型，进一步提高了该方法的定位性能。