报告人：吴毓朗博士

报告题目：Adaptive feedback Convolutional-neural-network-based High-resolution Reflection-waveform Inversion

摘要：Full-waveform inversion (FWI) applies non-linear optimization to estimate the velocity model by fitting the observed seismic data. With a smooth starting velocity model, FWI mainly inverts for the shallower background velocity model by fitting the observed direct, diving, and refracted data, and updates the interfaces by fitting the observed reflected data. As the deeper parts of the background velocity model cannot be effectively updated by fitting the reflected data in FWI, the deeper interfaces are less accurate than the shallower interfaces. In this presentation, I will introduce an adaptive feedback convolutional-neural-network-based high-resolution reflection-waveform inversion (CNN-RWI) to predict the model from the low-wavenumber initial model and the high-wavenumber migration image. Without the need for a representative training dataset, CNN-RWI iteratively updates CNN by the gradually more representative training data set, which is obtained from the latest more accurate CNN-predicted model by the proposed parcellation method. The more representative the training models are, the more accurate the CNN-predicted model becomes. The CNN-RWI can be easily extended to the more sophisticated wave equations (e.g., elastic and anisotropic wave equations) without suffering severe cross-talk issues. Synthetic examples using different portions of the Marmousi2 P-wave velocity model show that CNN-RWI inverts for both the shallower and deeper parts of models more accurately than the conventional FWI.

报告时间：2022年7月28日周四下午16:00-18:00

报告形式：腾讯会议：927-6491-6188

报告人简介：吴毓朗，2020年获得美国得克萨斯大学达拉斯分校地球科学博士学位， 2021年加入中科院地质与地球物理所油气资源研究院重点实验室从事博士后研究。研究方向为模型聚类与分割以及深度学习算法在地震反演中的应用。在地球物理权威期刊JGR - Solid Earth、Geophysics中发表多篇论文，其中2019年发表于Geophysics的论文（Parametric CNN-domain FWI）被美国勘探地球物理学会（SEG）选为亮点论文（Geophysical Bright Spots）。