Skip to Main content Skip to Navigation

Refinable resolution & precision for volume mesh compression & simulation in geosciences

Abstract : Research has entered a data-intensive era. The growing quantity of scientific information challenges several application domains. That happens notably for the simulation science, where huge data volumes create bottlenecks in high-performance computing workflows. Several data reduction methods are currently being developed as potential answersWe focus on geological modeling and reservoir simulation workflows. Geoscience models are composed of heterogeneous information, including mesh geometry and petrophysical properties, and may contain several millions of cells. Upscaling and upgridding techniques are commonly used for simulation time reduction. Yet there are often ad-hoc, and do not fully answer all the data manipulation needs: visualization, storage and ultimately well production prediction at the appropriate resolution and precision.We propose in this work a comprehensive methodology with refinable resolution and precision based on HexaShrink, a wavelet multiscale decomposition. We assess its performance on size reduction and visual relevance with lossless and lossy compression in a benchmark of meshes using entropy and zerotree coders.We also extensively test the impact of refinable resolution and precision on simulation. We specifically designed case-studies based on Lundi, a representative model of different geological environments. Results compare positively with state-of-the-art coders SZ and ZFP, by designing objective performance metrics that correlate well to subjection reservoir production assessment. Finally, we designed a complete workflow including a shareable mesh (Lundi, to be released as open data), designed for simulation in reservoir engineering. This model (meant to serve as a benchmark) can be processed by our multiscale decomposition, and then compressed by different encoders (lossless, or progressive/lossy) and compared to alternative compression methods. The resulting compressed meshes, generated at refinable resolution and precision, can be processed, and their quality can be evaluated at several steps of a simulation workflow.
Complete list of metadata
Contributor : Abes Star :  Contact Connect in order to contact the contributor
Submitted on : Thursday, July 15, 2021 - 10:49:16 AM
Last modification on : Monday, October 11, 2021 - 5:11:15 PM
Long-term archiving on: : Saturday, October 16, 2021 - 6:18:35 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03286798, version 1



Lauriane Bouard. Refinable resolution & precision for volume mesh compression & simulation in geosciences. Signal and Image processing. Université Côte d'Azur, 2021. English. ⟨NNT : 2021COAZ4014⟩. ⟨tel-03286798⟩



Record views


Files downloads