自适应稀疏反演多次波压制方法

Adaptive sparse inversion for multiples suppression

针对利用表面相关多次波压制方法(EPSI)人工拾取一次波同相轴来估计初始震源子波及其迭代更新可能存在的识别错误和拾取参数设置误差问题,提出了自适应稀疏反演多次波压制方法(AEPSI),解决了初始震源子波及震源子波迭代更新问题。该方法将EPSI过程表示为双线性算子一次波脉冲响应和震源子波的约束优化问题,用理论震源子波作为优化问题的初始点求取一次波脉冲响应。根据输入数据、一次波、多次波的期望能量关系,确定震源子波的初始振幅,并将求取地震子波问题转化为无约束非线性优化问题,采用置信域(trust region)优化算法进行迭代求解。在迭代过程中,震源子波的主频、振幅、相位等特性自适应地逐步逼近实际地震子波。模型和实际数据算例表明,自适应稀疏反演多次波压制方法正确有效、稳定收敛,对初始震源子波误差具有较高的容忍度,处理过程完全由数据驱动,不需要通过输入记录的多维互相关来估计初始震源子波,避免了求取一次波脉冲响应时的人工拾取和时窗设置。

In the estimation of primaries by sparse inversion(EPSI)method,the arrival of primary waves needs to be manually picked up to estimate the source wavelet.This may cause errors in identification and parameter setting.To address this issue,an adaptive estimation of primaries by sparse inversion(AEPSI)was proposed.In this method,the EPSI process was expressed as the primary impulse response of the double linear operator and constrained optimization of the source wavelet.Taking the theoretical source wavelet as the initial point of the optimization,the impulse response of the primary wave was calculated.Based on the energy relation among the input data,the primary wave and its multiples,as well as the original amplitude energy of the source wavelet were determined.The seismic wavelet calculation problem was then converted into an unconstrained nonlinear optimization problem,which was solved iteratively using the trust region optimization algorithm.During the iteration,the dominant frequency,amplitude,and phase of the source wavelet was able to approximate the practical seismic wavelet adaptively.This method does not require the multi-dimensional cross-correlation of the input records to estimate the initial source wavelet,and avoids the manual picking and time gate setting in calculating the primary impulse response.It has a high tolerance for errors in the initial source wavelet;moreover the processing is completely data driven.Tests on simulated and actual data showed that this method is effective and has a stable convergence.