Velocity inversion in cross-hole seismic tomography bycounter-propagation neural network, genetic algorithmand evolutionary programming techniques
Velocity inversion in cross-hole seismic tomography bycounter-propagation neural network, genetic algorithmand evolutionary programming techniques
复制标题
DOI：
10.1046/j.1365-246x.1999.00835.x
复制
发表时间：
1999-07
影响因子：
2.8
通讯作者：
S. Nath;S. Chakraborty;S. Singh;N. Ganguly
中科院分区：
文献类型：
--
作者：
S. Nath;S. Chakraborty;S. Singh;N. Ganguly
传统的地震层析射线追踪和基于计算的反演技术的缺点包括假设每个源-接收器对只有一条射线路径，不包括头波，计算时间长，以及难以在复杂的速度分布中找到射线路径。因此，利用互易原理和动态规划方法开发了一种光线跟踪算法。该鲁棒正演计算程序随后用于井间地震速度反演。地震透射层析成像可以看作是一个函数近似问题；也就是说，将旅行时间向量映射到速度向量。这属于模式分类问题的范畴，因此我们提出了一种用于地下层析成像的前向反传播神经网络（CPNN）技术。然而，神经网络的局限性在于其在日常解释中需要进行详尽的训练。由于有时从较差的初始模型中找到最优解是最终目标，因此在井间走时层析成像中也采用了模拟演化等全局优化和搜索技术。遗传算法（GA）、进化策略和进化规划（EP）是模拟进化研究的主要途径。因此，本研究的一部分涉及层析成像应用的GA和EP方案。在目前的模拟进化工作中，引入了一种新的称为“区域生长突变”的遗传算子来加快搜索过程。通过三个综合实例证明了纯前向CPNN、GA和EP方法的潜力。第一个模型的速度层析图通过CPNN和GA两种方案都能得到两个等速区域的对角线方向速度对比的可信图像，但EP方案不能完全成像模型。在第二种情况下，虽然GA和EP方案在均匀背景下生成了准确的断层层速度分布，但CPNN方案高估了断层的垂直位移。从第三个合成模型的ga构建层析图中可以很容易地识别出煤层中的五个空洞，但CPNN和EP方案无法复制该模型。这些方法的性能随后在印度西孟加拉邦Raniganj煤田Dhandadih煤矿的实际现场环境中进行了测试。利用这些算法对225条地震道进行首次到达时反演，得到了1.0 ~ 2.5 km s−1的纵波速度分布。利用CPNN、GA和EP方案，成功圈定了Jambad Top煤层中疑似巷道的低速带（1.0 km s−1）。
The disadvantages of conventional seismic tomographic ray tracing and inversion by calculus-based techniques include the assumption of a single ray path for each source–receiver pair, the non-inclusion of head waves, long computation times, and the difficulty in finding ray paths in a complicated velocity distribution. A ray-tracing algorithm is therefore developed using the reciprocity principle and dynamic programming approach. This robust forward calculation routine is subsequently used for the cross-hole seismic velocity inversion. Seismic transmission tomography can be considered to be a function approximation problem; that is, of mapping the traveltime vector to the velocity vector. This falls under the purview of pattern classification problems, so we propose a forward-only counter-propagation neural network (CPNN) technique for the tomographic imaging of the subsurface. The limitation of neural networks, however, lies in the requirement of exhaustive training for its use in routine interpretation. Since finding the optimal solution, sometimes from poor initial models, is the ultimate goal, global optimization and search techniques such as simulated evolution are also implemented in the cross-well traveltime tomography. Genetic algorithms (GA), evolution strategies and evolutionary programming (EP) are the main avenues of research in simulated evolution. Part of this investigation therefore deals with GA and EP schemes for tomographic applications. In the present work on simulated evolution, a new genetic operator called ‘region-growing mutation’ is introduced to speed up the search process. The potential of the forward-only CPNN, GA and EP methods is demonstrated in three synthetic examples. Velocity tomograms of the first model present plausible images of a diagonally orientated velocity contrast bounding two constant-velocity areas by both the CPNN and GA schemes, but the EP scheme could not image the model completely. In the second case, while GA and EP schemes generated an accurate velocity distribution of a faulted layer in a homogeneous background, the CPNN scheme overestimated the vertical displacement of the fault. One can easily identify five voids in a coal seam from the GA-constructed tomogram of the third synthetic model, but the CPNN and EP schemes could not replicate the model. The performances of these methods are subsequently tested in a real field setting at Dhandadih Colliery, Raniganj Coalfields, West Bengal, India. First arrival traveltime inversion by these algorithms from 225 seismic traces revealed a P-wave velocity distribution from 1.0 to 2.5 km s−1. A low-velocity zone (1.0 km s−1 ), the position of a suspected gallery in the Jambad Top coal seam, could be successfully delineated by CPNN, GA and EP schemes.