基于相似性重构低频数据的金属矿频域全波形反演

Full waveform inversion in the frequency domain of low-frequency seismic data based on similarity reconstruction for exploration of deep metallic ores

由于深部金属矿埋深和自身的复杂性,利用重、磁、电方法和一般的地震方法很难有效地对其进行高精度定位.全波形反演通过最小化模拟数据与观测数据的差异使深部金属矿的高精度探测成为可能,但全波形反演是一个局部优化过程,需要准确的低频数据作为起始,而这在一般的地震数据采集中难以做到.本文先在频域中使用伴随状态震源函数反演方法,通过震源附近的直达波能精确地反演出震源函数的形态.然后利用得到的高精度震源子波结合褶积与反褶积思想及相似性现象重构含有低频成分的自激自收数据.将该数据应用到全波形反演中,有效缓减了反演过程中出现的周波跳跃现象,并提高了模型反演的正确性.Marmousi模型和金属矿模型的数值模拟实验证明了新方法改善了在没有低频数据时的全波形反演结果,并有较好抗噪性.

Because burial depth and complexity,gravity,magnetic electrical and conventional seismic methods cannot accurately locate metallic ore bodies in the deep subsurface.To fill this gap,Full Waveform Inversion(FWI)makes it possible to image these geologic bodies at high precision by minimizing the misfit between simulated and measured data.However,the FWI method depends highly on the initial velocity model and the low-frequency seismic data,which is difficult in conventional seismic data acquisition.To solve this problem,this work uses the direct wave information for the adjoint-state seismic wavelet inversion to establish an accurate seismic wavelet for the FWI method.Combining convolution,de-convolution and similarity phenomenon,the low-frequency components in the self-exciting and self-receiving data can be effectively reconstructed,which reflects the macro structures of the velocity models.Numerical tests on the Marmousi and metal mine models show that the reconstructed low-frequency components in the self-exciting and self-receiving data can significantly improve the inversion results,even if the seismic data lacks frequency components below 7 Hz or the seismic data contaminated by strong noise.