Towards automatically building starting models for full-waveform inversion using global optimization methods: A PSO approach via DEAP + Devito

Abstract

In this work, we illustrate an example of estimating the macromodel of velocities in the subsurface through the use of global optimization methods (GOMs). The optimization problem is solved using DEAP (Distributed Evolutionary Algorithms in Python) and Devito, python frameworks for evolutionary and automated finite difference computations, respectively. We implement a Particle swarm optimization (PSO) with an "elitism strategy" on top of DEAP, leveraging its transparent, simple and coherent environment for implementing of evolutionary algorithms (EAs). The high computational effort, due to the huge number of cost function evaluations (each one demanding a forward modeling step) required by PSO, is alleviated through the use of Devito as well as through parallelization with Dask. The combined use of these frameworks yields not only an efficient way of providing acoustic macro models of the P-wave velocity field (V p ), but also significantly reduces the amount of human effort in fulfilling this task.

GPU Support for Automatic Generation of Finite-Differences Stencil Kernels

“…The open-source project Devito R [23,1] has been attracting the attention of academic [24,25] and industrial [26] community. As a DSL for seismic inversion applications, it already provides a set of automated performance optimizations during code generation that allow user applications to fully utilize the target hardware without changing the model specification, such as vectorization, shared-memory parallelism, loop blocking, auto-tuning, common sub-expression elimination, cross-iteration redundancy elimination (CIRE), expression hoisting and factorization.…”

Full waveform inversion based on dynamic data matching of convolutional wavefields

“…Huang et al (2021) used the time-warping function as the extension in the data space to solve the velocity model and time-warping extension in a single optimization problem by the alternate direction method. Some other solutions, such as gradient sampling (Yang et al, 2020), wavefield reconstruction (Rizzuti et al, 2021), Bayesian non-linear inversion (Guo et al, 2020), wide-angle seismic acquisition (Guo et al, 2022), and global optimization (Mojica and Kukreja, 2019), can also mitigate cycle skipping.…”

Fractal-Constrained Crosshole/Borehole-to-Surface Full-Waveform Inversion for Hydrogeological Applications Using Ground-Penetrating Radar

“…Moreover, in an effort to reduce the computational requirements, 2.5D forward solvers and 2D to 3D transformations have been proposed in order to replace costly 3D simulations [34]. Furthermore, the presence of numerous local minimal in the optimization space has been tackled via ray-based initial models [16], using global optimizers [35], [36], and by gradually increasing the selected bandwidth as the optimization progresses [37]. Lastly, a further increase in resolution was made possible by addressing sharp boundaries and promoting "blocky" solutions using total variation regularization constraints [38].…”