Wednesday 19 March 2025
A new approach has been developed to tackle the complex problem of simulating seismic wavefields in complex media, such as those found in Earth’s crust. Seismic wavefields are crucial for understanding earthquakes and oil exploration, but traditional methods can be computationally expensive and limited in their accuracy.
The team behind this innovation used a combination of machine learning and physics to create a framework that can efficiently simulate wavefields in varying frequencies and velocities. They achieved this by introducing a novel neural network architecture that incorporates the fundamental laws of physics into its design.
This approach, known as Meta-LRPINN, uses a technique called singular value decomposition (SVD) to decompose the weights of the neural network. This reduces the number of parameters needed, making the model more efficient and easier to train. Additionally, the team developed an innovative frequency embedding hypernetwork that links input frequencies with the singular values in the SVD decomposition.
The researchers tested their framework on a range of scenarios, including layered velocity models, which are commonly used in seismic imaging. They found that Meta-LRPINN was able to accurately simulate wavefields across multiple frequencies and velocities, outperforming traditional methods in terms of speed and accuracy.
One of the key advantages of this approach is its ability to adapt to different frequency ranges and velocity distributions. This means that it can be used for a wide range of applications, from earthquake simulation to oil exploration.
The team’s findings have significant implications for the field of seismic imaging. By providing an efficient and accurate way to simulate wavefields, Meta-LRPINN has the potential to revolutionize our understanding of complex media and improve our ability to extract valuable resources from beneath the Earth’s surface.
In the future, the researchers plan to continue refining their framework and exploring its applications in various fields. They also hope to collaborate with other experts in the field to further develop this innovative technology.
Cite this article: “Simulating Seismic Wavefields with Meta-LRPINN: A Novel Framework for Complex Media”, The Science Archive, 2025.
Seismic Wavefields, Machine Learning, Physics, Neural Networks, Meta-Lrpinn, Singular Value Decomposition, Frequency Embedding, Hypernetworks, Seismic Imaging, Earth’S Crust.
