基于复频率衰减反传算子的全波形反演梯度构建

从梯度构建中反传算子的角度出发，本文分析了影响梯度精度的因素， 并提出了一种新的基于复频率衰减反传算子的全波形反演梯度构建法， 以压制梯度构建过程中产生的弧状噪音；针对全波形反演的巨大存储量和精度问题，本文采用限内存的BFGS(L-BFGS)算法进行迭代反演．对平层模型、逆掩断层模型以及凹陷模型的测试结果表明，与常规梯度算法相比，基于复频率衰减反传算子算法所构建梯度的精度得到明显改善．用复杂断块模型进行全波形反演测试，并将其与常规算法得到的反演速度精度进行对比，结果验证了该方法的正确性及其对提高梯度精度的有效性． 

Full waveform inversion gradient construction based on the complex frequency attenuation back propagation operator

This paper analyzed the factors affecting the gradient accuracy from the view of back propagation operator, and proposed a new method for gradient construction of full waveform inversion based on complex frequency attenuation back propagation operator. The gradient construction noise can be suppressed with the proposed method. During the inversion, the BFGS (L-BFGS) method is adopted to balance the relation between the huge storage capacity and the accuracy. The tests based on single layer model, overthrust fault model and sag model validate that the proposed method can suppress the noise in the gradient construction comparing with the conventional gradient method. Then the full waveform inversion is tested on a complex faulted blocks model, and the inversion accuracy is compared with that of the conventional method, proving that the proposed method in this paper is correct and effective in improving the gradient accuracy.