Computational wave imaging (CWI) aims to reveal hidden physical properties of a medium, such as internal density and bulk modulus, by analyzing wave signals that pass through it. This technique has diverse applications, including seismic full-waveform inversion for subsurface exploration, ultrasound computed tomography in medicine, and acoustic imaging in materials science. However, CWI presents a challenging mathematical problem due to its ill-posed nature. In this tutorial, we will explore how integrating machine learning (ML) with fundamental physics can address these challenges and improve solution efficiency. This integration can lead to more accurate, faster, and enhanced wave imaging techniques. We will introduce advanced ML models, such as diffusion models, to tackle CWI problems effectively. Furthermore, we will present case studies illustrating the application of CWI and ML methods in monitoring carbon sequestration leakage, a critical real-world problem.
[1]. Y. Lin, J. Theiler and B. Wohlberg, “Physics-Guided Data-Driven Seismic Inversion: Recent Progress and Future Opportunities in Full Waveform Inversion,” IEEE Signal Processing Magazine, 40(1), pp. 115-133, 2023.
[2]. Y. Wu and Y. Lin, InversionNet: An Efficient and Accurate Data-driven Full Waveform Inversion, IEEE Transactions on Computational Imaging, 6(1):419-433, 2019.
[3]. C. Deng, S. Feng, H. Wang, X. Zhang, P. Jin, Y. Feng, Q. Zeng, Y. Chen and Y. Lin, “OpenFWI: Large-Scale Multi-Structural Benchmark Datasets for Seismic Full Waveform Inversion,” the Conference on Neural Information Processing Systems (NeurIPS), 2022.
[1] Z. Zhou, Y. Lin, Z. Zhang, Y. Wu, Z. Wang, R. Dilmore, and G. Guthrie, A Data-Driven CO2 Leakage Detection Using Seismic Data and Spatial-Temporal Densely Connected Convolutional Neural Networks, International Journal of Greenhouse Gas Control, Vol 90, 2019.
[2] J. Yang, H. Wang, Y. Sheng, Y. Lin, and L. Yang, A Physics-guided Generative AI Toolkit for Geophysical Monitoring. ArXiv. /abs/2401.03131, 2024 (Accepted to DAC 2024).