Acoustic tomography is an important test method in engineering geology and underground engineering, however, the problems of model parameterization, forward modeling and inversion in the process of tomographic imaging still need further study. In this work, the triangulation method of two-dimensional complex area was studied and the global algorithm of triangular ray tracing in two-dimensional complex area was formulated. Numerical simulation results of triangular acoustic ray tomography show that with the increase of iteration times, the ray tracing becomes increasingly close to the ray-tracing forward from the velocity model; through seven iterations, the low-speed plate with a thickness of only 0.56 m can also be clearly reflected. Besides, based on triangular ray-tracing tomography of the two-dimensional complex structure, the deep acoustic detection of coal seam floor failure zone is carried out. After the initial velocity model is determined by the single-hole acoustic detection results, the velocity distribution of the floor between holes is obtained by triangular network acoustic tomography from the premining, mid-mining and postmining acoustic penetration detection. Field measurements show that the 17-m-span acoustic wave penetration is successfully realized in sandstone and mudstone strata of coal seam floor, with accurate delineation of the development range of water-conducting fissures in floor failure. The triangular acoustic tomography method makes full use of the characteristics of flexible model division and makes the imaging results appear closer to the actual structural morphological characteristics.

中文翻译：

---

基于三角网络声学CT层析成像的煤层顶板破坏深度反演及应用研究

声层析成像技术是工程地质学和地下工程中的一种重要测试手段，然而，层析成像过程中模型参数化，正演建模和反演等问题仍需进一步研究。本文研究了二维复杂区域的三角剖分方法，提出了二维复杂区域三角射线跟踪的全局算法。三角声波层析成像的数值模拟结果表明，随着迭代时间的增加，射线追踪变得越来越接近速度模型中的射线追踪。通过七次迭代，厚度仅为0.56 m的低速板也可以清晰地反映出来。此外，基于二维复杂结构的三角射线追踪层析成像，进行了煤层底板破坏带的深层探测。在通过单孔声波检测结果确定初始速度模型后，通过三角网声层析成像从开采前，开采中和开采后的声波穿透检测获得孔之间地板的速度分布。现场测量表明，在煤层底板的砂岩和泥岩地层中成功实现了17米跨度的声波穿透，并准确描述了底板破裂时导水裂缝的发展范围。三角声层析成像方法充分利用了柔性模型划分的特征，使成像结果看起来更接近实际的结构形态特征。