TY - JOUR

T1 - A new full waveform inversion method based on shifted correlation of the envelope and its implementation based on OPENCL

AU - Wu, Zedong

AU - Alkhalifah, Tariq Ali

AU - Zhang, Zhendong

AU - Alonaizi, Faisal

AU - Almalki, Majed

N1 - KAUST Repository Item: Exported on 2020-10-01
Acknowledgements: We thank KAUST an KACST for their support. We thank the SWAG group for collaborative environment. The numerical experiments were performed on the super computer SANAM, which is hosted by KACST, Saudi Arabia. We also thank the support team of SANAM for their useful help. Our special thanks are extended to the associate editor Sjoerd de Ridder and two anonymous reviewers for their fruitful suggestions and comments.

PY - 2019/5/1

Y1 - 2019/5/1

N2 - Standard full waveform inversion (FWI) attempts to minimize the difference between observed and modeled data. When the initial velocity is kinematically accurate, FWI often converges to the best velocity model, usually of a high-resolution nature. However, when the modeled data using an initial velocity is far from the observed data, conventional local gradient based methods converge to a solution near the initial velocity instead of the global minimum. This is known as the cycle-skipping problem, which results in a zero correlation when observed and modeled data are not correlated. To reduce the cycle-skipping problem, we compare the envelope of the modeled and observed data instead of comparing the modeled and observed data directly. However, if the initial velocity is not sufficient, the correlation of the envelope of the modeled and observed data might still be zero. To mitigate this issue, we propose to maximize both the zero-lag correlation of the envelope and the non-zero-lag correlations of the envelope. A weighting function with maximum value at zero lag and decays away from zero lag is introduced to balance the role of the lags. The resulting objective function is less sensitive to the choice of the maximum lag allowed and has a wider region of convergence with respect to standard FWI and envelope inversions. The implementation has the same computational complexity as conventional FWI as the only difference in the calculation is related to a modified adjoint source. The implementation of this algorithm was performed on an AMD GPU platform based on OPENCL and provided a 14 times speedup over a CPU implementation based on OPENMP. Several numerical examples are shown to demonstrate the proper convergence of the proposed method. Application to the Marmousi model shows that this method converges starting with a linearly increasing velocity model, even with data free of frequencies below 3 Hz.

AB - Standard full waveform inversion (FWI) attempts to minimize the difference between observed and modeled data. When the initial velocity is kinematically accurate, FWI often converges to the best velocity model, usually of a high-resolution nature. However, when the modeled data using an initial velocity is far from the observed data, conventional local gradient based methods converge to a solution near the initial velocity instead of the global minimum. This is known as the cycle-skipping problem, which results in a zero correlation when observed and modeled data are not correlated. To reduce the cycle-skipping problem, we compare the envelope of the modeled and observed data instead of comparing the modeled and observed data directly. However, if the initial velocity is not sufficient, the correlation of the envelope of the modeled and observed data might still be zero. To mitigate this issue, we propose to maximize both the zero-lag correlation of the envelope and the non-zero-lag correlations of the envelope. A weighting function with maximum value at zero lag and decays away from zero lag is introduced to balance the role of the lags. The resulting objective function is less sensitive to the choice of the maximum lag allowed and has a wider region of convergence with respect to standard FWI and envelope inversions. The implementation has the same computational complexity as conventional FWI as the only difference in the calculation is related to a modified adjoint source. The implementation of this algorithm was performed on an AMD GPU platform based on OPENCL and provided a 14 times speedup over a CPU implementation based on OPENMP. Several numerical examples are shown to demonstrate the proper convergence of the proposed method. Application to the Marmousi model shows that this method converges starting with a linearly increasing velocity model, even with data free of frequencies below 3 Hz.

UR - http://hdl.handle.net/10754/652849

UR - https://www.sciencedirect.com/science/article/pii/S0098300418303753

UR - http://www.scopus.com/inward/record.url?scp=85065490218&partnerID=8YFLogxK

U2 - 10.1016/j.cageo.2019.04.007

DO - 10.1016/j.cageo.2019.04.007

M3 - Article

SN - 0098-3004

VL - 129

SP - 1

EP - 11

JO - Computers & Geosciences

JF - Computers & Geosciences

ER -