基于高阶振幅信息的时间域全波形反演

Time-domain full waveform inversion based on high-order amplitude information

常规全波形反演的目标函数具有强非线性特点,当速度反演的初始值与真实值相差较大时,目标函数将会陷入局部极小值,无法获得速度模型的高精度反演结果.为了解决这个问题,可以通过降低地震数据的复杂性,来缓解全波形反演目标函数的非线性程度.本文提出基于高阶振幅信息的时间域全波形反演方法,通过提取复地震信号,在相位不变的情况下提高振幅信息的阶数,成功降低目标函数的复杂度,进而缓解全波形反演的周期跳跃问题.Marmousi模型测试表明,基于高阶振幅信息的时间域全波形反演方法可以获得一个较好的初始速度模型,进而提高全波形反演的速度建模精度.

The objective function of the conventional Full Waveform Inversion(FWI)has strong nonlinear characteristics.When the initial value of velocity inversion is far from the true value,the objective function will fall into local minima,and high-resolution inversion results of the velocity model cannot be obtained.To solve this problem,the nonlinearity of the FWI objective function can be mitigated by reducing the complexity of seismic data.In this paper,a time-domain FWI based on high-order amplitude information is proposed.The different orders of amplitude information can be obtained by extracting complex seismic signals with phase unchanged.In this way,the complexity of the objective function is successfully reduced,and the cycle skipping of FWI is then mitigated.Marmousi model tests show that the time-domain FWI based on high-order amplitude information can obtain a better initial velocity model for FWI to improve the resolution of the velocity model.