Periodic Reporting for period 1 - TerraPINN (Toward fully physics based probabilistic seismic hazard assessment using physics informed neural networks)
Periodo di rendicontazione: 2022-03-01 al 2024-02-29
The ability to accurately simulate seismic wavefields is an essential part of our ability to understand seismic hazard and it's potential risks to life and built infrastructure. It is also a key component in our ability to image Earth's subsurface, which has both innate scientific interest and is of societal importance due to our reliance on Earth resources. Unfortunately, seismic wavefield simulation is extremely computationally expensive, often requiring the resources of entire supercomputer facilities to run, which hinders our ability to use seismic information. The objective of this project was to investigate machine learning based methods for accelerating seismic wavefield computations, in particular for ground motion studies of seismic hazard. We chose to investigate the recently developed physics-informed neural-network (PINN) approach, as unlike traditional machine learning methods, PINNs do not require reference data for machine learning, instead relying on our knowledge of underlying physical principles.
We worked on the technical development of PINN methods for seismic wavefield computation. In particular, it has become well known since the original development of PINNs in 2019 that they perform very well on "smooth" problems, where as "rough" problems (those with high spatial frequency content) can pose challenges in training. Seismic wavefields are relatively rough as they contain broadband temporal and spatial frequency content. However, data-driven machine learning methods (the vast majority of AI/ML platforms currently extant) can capture high degrees of spatial complexity, at the cost of requiring training data. We recognized that the complexity of the seismic wavefield is primarily in the direction pointing directly away from a seismic source, but that the azimuthal direction is relatively simple (imagine cutting vertically through the center of an orange to take out a wedge; any one wedge looks quite complicated, but if you were to chose another wedge to cut, the two wedges would look very similar - this is azimuthal symmetry). As such, we can train a data-driven machine learning model on a lower-dimensional problem that is simple but still captures the spatial frequency content, and then use a PINN to then solve the more complicated higher dimensional problem using the data-driven solution as a reference. The result is that the PINN does not need to learn complex spatial patterns itself, but rather warps the underlying complexity from the data driven model, thereby avoiding the issue that PINNs have in learning high frequency content. We have reported these results within leading scientific conferences, including in a talk at the American Geophysical Union Fall Meeting 2023, and a publication is in preparation. Performing this research further inspired 3 scientific publications during the project, 2 of which are published in the diamond open access journal Seismica, and one in Geophysical Journal International as a gold open access paper.
The hybrid axisymmetric PINN wavefield solver improved the efficiency of the previous state-of-the-art PINN for seismic problems by two orders of magnitude, which will enable further scale-up to complex seismic problems. This has the potential to significantly improve our ability to rapidly model physics based ground motion modelling and seismic tomography, which will improve our ability to mitigate seismic hazard and improve our utilization of Earth resources respectively. With the support of this EU funded project, the lead researcher also supervised three research students on the topics of soil seismology, other alternative PINN methods for seismic wavefields, and AI/ML seismic imaging methods for carbon sequestration, further building societal capacity for important Earth science research.