一种用于背景速度估计的新全局优化算法
|
摘要
利用餐前地震数据正确地估计地下介质的背景速度对于复杂构造成像具有重要意义。这一问题的复杂性需要一种良好的优化方法来求解。本文给出了一种新的用于求解连续全局优化问题的计算方法，并将其应用于非线性速度反演计算。与随机法（如模拟退火、遗传算法、均场退火等）不同的是，新的优化方法利用一动态方程来控制寻优计算过程。该动态方程的优点是利用梯度信息使解快速逼近极小点；当解逼近当前极小点后，动态方程的第二项可以使解逃逸局部极值，使其具有全局寻优能力；梯度信息的利用，使求解效率大大提高。这种基于优化算法的速度估计方法在计算目标函数时勿需大量的反射波旅行时拾取，所需要的只是从叠加剖面中得到的零偏移距旅行时，利用相似度准则建立优化目标函数。采用上述寻优方法，自动完成背景速度估算。文中给出算例，证明了方法的计算效率较随机法有明显提高，速度估计算例也给出了正确的结果。
To derive correct medium velocity from prestack seismic data is very significantto complicated structure imaging. The complexity solution of the above problemneeds a good optimization method. A new successive overall optimization calculation method will be used in the solution and non--linear velocity inversion calculation. The method is different from random methods (such as analog annealing, genetic algorithm and average field annealing), and it applies a dynamic equation tocontrol optimizing computation process. The dynamic equation uses gradient information to make the solution fast approximate the minimum point. The 2nd term ofthe dynamic equation may make the solution escape local extreme value and causethe overall optimization capacity when the solution approximates the minimumpoint; the application of gradient information has made the solution efficiency improved obviously. There will be no need to pick large amount of reflection traveltimes in calculation of objective function by the velocity estimation method based onoptimization algorithm. What is needed is only to derive zero--offset travel-timesfrom stack section to establish optimization objective function with the use of similitude criterion. Background velocity estimation will be automatically completed bythe method. The example in this paper proves that its computing efficiency is muchbetter than that of random method. Correct result has been obtained in velocity estimation.
To derive correct medium velocity from prestack seismic data is very significantto complicated structure imaging. The complexity solution of the above problemneeds a good optimization method. A new successive overall optimization calculation method will be used in the solution and non--linear velocity inversion calculation. The method is different from random methods (such as analog annealing, genetic algorithm and average field annealing), and it applies a dynamic equation tocontrol optimizing computation process. The dynamic equation uses gradient information to make the solution fast approximate the minimum point. The 2nd term ofthe dynamic equation may make the solution escape local extreme value and causethe overall optimization capacity when the solution approximates the minimumpoint; the application of gradient information has made the solution efficiency improved obviously. There will be no need to pick large amount of reflection traveltimes in calculation of objective function by the velocity estimation method based onoptimization algorithm. What is needed is only to derive zero--offset travel-timesfrom stack section to establish optimization objective function with the use of similitude criterion. Background velocity estimation will be automatically completed bythe method. The example in this paper proves that its computing efficiency is muchbetter than that of random method. Correct result has been obtained in velocity estimation.
引文
1Al-Yahya.Velocity analysis by iterative profile migration Geophysics ,1989,54(6):718～729
2 Macky S and Abma R.Imaging and velocity estimation with depth focusing analysis.Geophysics,1992,5:(12):1608～1622
3Landa E,Beydoun W and Tarantola A Reference velocity model estimation from prestacked waveforms;Coherency optimization by simulaling Geophysics.1989,54(8):984～990
4刘勇.非线性并行算法(第二册),科学出版社,1996
5Ingbeer L.Very fast simulated reannealing Math Comput Modeling 1989,12:967～973
6Shang Y and Wah B W.Global optimization for neural network training Computer,1996,45～54
7Rothman D H. Automatic estimstion of large residual statics correction Geophysics , 1986, 51( 2): 332～ 346
8Sambridge M and Drijkoningen G.Genetic algorithms in seismic waveform inversion Geophsics J Inter-nat,1992,109:323～343
9Sen M K and Stoffa P L.Nonlinear one dimensional seismic waveform inversion using simulated anneal-ing.Geophysics,1991,56(10):1624～1638
10长春,李幼铭,冷传波.适用于复杂介质中的3-D射线跟踪算法.地球物理学报,1996,39(4)
2 Macky S and Abma R.Imaging and velocity estimation with depth focusing analysis.Geophysics,1992,5:(12):1608～1622
3Landa E,Beydoun W and Tarantola A Reference velocity model estimation from prestacked waveforms;Coherency optimization by simulaling Geophysics.1989,54(8):984～990
4刘勇.非线性并行算法(第二册),科学出版社,1996
5Ingbeer L.Very fast simulated reannealing Math Comput Modeling 1989,12:967～973
6Shang Y and Wah B W.Global optimization for neural network training Computer,1996,45～54
7Rothman D H. Automatic estimstion of large residual statics correction Geophysics , 1986, 51( 2): 332～ 346
8Sambridge M and Drijkoningen G.Genetic algorithms in seismic waveform inversion Geophsics J Inter-nat,1992,109:323～343
9Sen M K and Stoffa P L.Nonlinear one dimensional seismic waveform inversion using simulated anneal-ing.Geophysics,1991,56(10):1624～1638
10长春,李幼铭,冷传波.适用于复杂介质中的3-D射线跟踪算法.地球物理学报,1996,39(4)
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心