论文部分内容阅读
地球物理资料反演中通常采用局部和全局最优化算法。每一种算法都有其固有的优点和缺点。本文提出把两种算法组合起来的几种混合方法,以便克服它们的缺点,并且利用两种算法的突出特点。实际上,我们是把局部共轭梯度(CG)算法与全局快速模拟退火(VFSA)方法结合起来,解决地球物理重大问题的。我们开展了系统性研究以寻找组合CG和VFSA算法的最有效策略。并且为进一步研究推荐两种方法。 首先在一组野外一维Schlumberger电阻率测深资料上进行七种不同混合算法的试验,试验的结果与单独基因算法(GA)、模拟退火法和局部搜索算法的结果进行了比较。可以发现几乎所有已提出的混合算法在计算上都比传统全局最优化方法更有效。在找到最有效的混合算法后,我们把它们应用于偏移距——时间域地震记录的速度分析中。最后,我们把混合算法应用到亚利桑那州Safford的浸染硫化带上采集到的二维野外电阻率剖面上,并将混合反演结果与已发表的结果进行了比较。
Geophysical data inversion usually adopts local and global optimization algorithms. Each algorithm has its inherent advantages and disadvantages. In this paper, we propose several hybrid methods combining the two algorithms in order to overcome their shortcomings and make use of the salient features of the two algorithms. In fact, we combine the local conjugate gradient (CG) algorithm with the global fast simulated annealing (VFSA) method to solve the major problems of geophysics. We conducted a systematic study to find the most effective strategy for combining CG and VFSA algorithms. And two methods are recommended for further research. Seven different hybrid algorithms were first tested on a set of field-based Schlumberger resistivity sounding data. The results of the experiments were compared with the results of GA, simulated annealing and local search algorithms. It can be found that almost all proposed hybrid algorithms are computationally more efficient than traditional global optimization methods. After finding the most efficient hybrid algorithms, we apply them to the velocity analysis of offset-time domain seismograms. Finally, we applied the hybrid algorithm to the two-dimensional field resistivity profile collected on dip-sulphide zones in Safford, Arizona, and compared the results of the hybrid inversion with published results.