Xinru Mu, Shijun Cheng, and Tariq Alkhalifah

 GitHub: https://github.com/DeepWave-KAUST/Separation-PINN-pub

A physics-informed neural network (PINN) approach for P- and S-wave separation to enhance multi-component seismic data processing. Unlike traditional supervised learning methods, which require a large amount of labeled data, our method is unsupervised and applicable to unlabeled data. It effectively captures multi-scale wavefield features and provides a mesh-free numerical solution. Numerical tests demonstrate its accuracy in both homogeneous and heterogeneous media.