Quick Start

Welcome to Seistorch! This quick start guide will walk you through the basics of using Seistorch for seismic simulations and Full-Waveform Inversion (FWI).

Note: The FWI results demonstrated in our example are approximate and indicate incomplete inversion. To achieve better inversion results, you may need to adjust various parameters, such as the learning rate, decay rates, source wavelet frequencies, and potentially other hyperparameters. Fine-tuning these parameters can significantly impact the quality and convergence of the inversion process, allowing you to obtain more accurate subsurface models. It often requires experimentation and tuning to find the optimal set of parameters for your specific seismic data and geological conditions.

We’ll cover the following topics:

1. Forward modeling: Simulate seismic wave propagation in 2D or 3D space.

2. Full waveform inversion: This topic covers folowing fwi examples:

   • source-encoding-based fwi in acoustic and elastic cases

   • batched and mpi-based tradtional fwi

   • memory-saving fwi using boundary-saving

   • towed acquistion fwi with marmousi model

   • inversion with multiples

3. Reverse time migration: This topic introduce how to perform reverse time migration using seistorch.

4. FWI combined with Neural network: This topic will introduce some examples of fwi combied with tradtional neural network methods.

5. Other usage: This topic covers following examples:

   • how to perform source inversion

   • how to calculate the adjoint source using torch