摘要：
In the disaster geological problems such as shallow ground fissure etc.the demand for fine detection has been increasing in recent years.The first-break velocity tomography technique can provide a high accuracy of the vertical and horizontal anomalies of the shallow velocity field,and provide the basis for identifying the disaster geological phenomena such as ground fissures etc.In general,under certain conditions of acquisition parameters the accuracy of inversion imaging is limited by the algorithm,the grid type and the scale of the grid.Therefore,in order to improve the accuracy of the calculation and keep the grid type unchanged,the inversion of the grid need to be increased.On the basis of the improved algorithm,the effect of incremental and non-deceleration and high precision can be realized.In this paper,OpenMP-based stand-alone multi-core parallel technology is used,and the first-break tomography parallel algorithm is implemented.In the forward part,the improved traveltime linear interpolation method is used.The method of column calculation is used,and the method of line calculation is also used.In this case,the rays of each direction are taken into account,so that the minimum traveltime is obtained.In the inversion section,the LSQR method and the regularization technique are used.The results of theoretical and practical data test show that the convergence rate is fast and the inversion result is better.At the same time,compared with the conventional serial algorithm,the speed and efficiency of the algorithm are greatly improved under the condition of added sampling and reduced grid scale.Compared with commercial software,the change of computing efficiency is not obvious,but the accuracy and distinguishability of inversion results are better.%随着浅层地裂缝等地质灾害问题精细探测需求的增大,采用初至波速度层析反演技术,可以提供较高精度的浅层速度场的纵横向异常变化,可为识别地裂缝等地质灾害现象提供依据.通常,在采集参数一定的条件下,其反演成像精度受算法、网格类型和剖分尺度的限制,在网格类型不变的条件下,要提高精度就必须加密正反演计算网格,改进算法,从而实现增量而不减速且高精度的效果.文中利用OpenMP的单机多核并行技术,探讨并实现了初至波层析成像并行算法,其中正演部分使用了改进的旅行时线性插值法,针对原有的按列计算,加入了按行计算,充分考虑到来自各方向的射线,使得计算出的旅行时最小,反演部分使用了能迅速稳定收敛的LSQR法以及正则化技术,通过理论和实际资料测试,其收敛速度快,反演结果较好.同时,在加密采样和缩小网格尺度的条件下,与常规串行算法相比,其运算速度和效率有较大提高,与商业软件比较,其运算效率差异不大,但反演结果的精度和可分辨性较好.
