Automated decision making and guided QC for full-waveform inversion

Automation can improve turnaround time for full-waveform inversion (FWI) projects, through reducing the burden on the user to select parameters, making the process more robust, and accelerating decision making.

We begin by considering a previous approach that generates automatic quality control assessments (QCs), internal to the FWI process, that are used to manage decisions on the parameterization of the FWI workflow, without the need for user intervention.

This automatic internal QC workflow incorporates an FWI objective function that automatically balances dynamic and kinematic contributions. While the internal QCs are used for automatic control, we show how the attributes generated during the internal QC workflow can be aggregated and reported to provide additional external QCs that can be exposed to the user, providing insight on how the inversion is proceeding and can inform decisions that need to be made.

The external QCs can introduce a significant amount of additional data to be analysed. To address this, we introduce an unsupervised learning workflow, that allows for guided analysis of these QCs, which can allow for faster (and ultimately more automated) decision making in FWI.

This data-science based QC workflow is integrated as part of a cloud-native FWI solution.