Chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

Lucia, Marco De; Kühn, Michael

doi:10.31223/x5g30s

Cited by 2 publications

(4 citation statements)

References 20 publications

(33 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DecTree v1.0 is a model experiment set up and evaluated in the R environment. All the code used in the present work is available under the LGPL v2.1 licence at Zenodo with DOI https://doi.org/10.5281/zenodo.4569573 (De Lucia, 2021a).…”

Section: Discussionmentioning

confidence: 99%

DecTree v1.0 – chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

Lucia

Kühn

2021

Geosci. Model Dev.

Self Cite

View full text Add to dashboard Cite

Abstract. The computational costs associated with coupled reactive transport simulations are mostly due to the chemical subsystem: replacing it with a pre-trained statistical surrogate is a promising strategy to achieve decisive speedups at the price of small accuracy losses and thus to extend the scale of problems which can be handled. We introduce a hierarchical coupling scheme in which “full-physics” equation-based geochemical simulations are partially replaced by surrogates. Errors in mass balance resulting from multivariate surrogate predictions effectively assess the accuracy of multivariate regressions at runtime: inaccurate surrogate predictions are rejected and the more expensive equation-based simulations are run instead. Gradient boosting regressors such as XGBoost, not requiring data standardization and being able to handle Tweedie distributions, proved to be a suitable emulator. Finally, we devise a surrogate approach based on geochemical knowledge, which overcomes the issue of robustness when encountering previously unseen data and which can serve as a basis for further development of hybrid physics–AI modelling.

show abstract

Section: Discussionmentioning

confidence: 99%

DecTree v1.0 – chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

Lucia

Kühn

2021

Geosci. Model Dev.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Leal et al [16] and De Lucia and Kühn [2] propose to replace the emulated geochemical prediction by a full calculation online during the RT simulation, if a quality criterion is not satisfied. Implementing this interesting possibility within our framework would be straightforward.…”

Section: Discussionmentioning

confidence: 99%

“…Leal et al [16] base it upon acceptable changes in chemical potential which for a chemical solver based on the law of mass actions (as Phreeqc in HPx) could be replaced by the activity or apparent chemical potential as outlined in Leal et al [15]. De Lucia and Kühn [2] resort to mass balance error but with our implementation there is by design no mass balance error. Indeed, since for each time step and grid node we emulate the aqueous concentrations from the total components' amounts and then re-calculate the solid amounts by subtracting the new aqueous concentrations from the total amounts, mass balance is always honored.…”

Section: Discussionmentioning

confidence: 99%

“…Prasianakis et al [21] used NNs to either predict permeability from porosity n a 1D problem or saturation index from the concentration of two components in a large 2D problem. Lastly, in a similar spirit of Leal et al [16], De Lucia and Kühn [2] present a strategy in which at runtime, full physics geochemical simulations are performed only if surrogate predictions are deemed implausible.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Speeding up reactive transport simulations in cement systems by surrogate geochemical modeling: deep neural networks and k-nearest neighbors

Laloy¹,

Jacques²

2021

Preprint

View full text Add to dashboard Cite

This study investigates how reactive transport (RT) simulation can be accelerated by replacing the geochemical solver the RT code by a surrogate model or emulator, considering either a trained deep neural network (DNN) or a k-nearest neighbor (kNN) regressor. We focus on 2D leaching of hardened cement paste under diffusive or advective-dispersive transport conditions, a solid solution representation of the calcium silicate hydrates and either 4 or 7 chemical components, and use the HPx reactive transport code as baseline. We find that after training, both our DNN-based and kNN-based codes, called HPx py -DNN and HPx py -kNN, can make quite accurate predictions for the simpler 4-components cement system while providing either a 4.2 to 6.8 (DNN) or 2.8 to 5.0 (kNN) speedup factor compared to HPx with parallelized geochemical calculations over 4 cores. Benchmarking against single-threaded HPx, these speedup factors become 16.2 to 24.5 and 10.8 to 18.0 for HPx py -DNN and HPx py -kNN, respectively. For the more complex 7-components cement system, no emulator that is globally accurate over the full space of possible geochemical conditions could be devised, neither using DNNs nor kNN. Instead we therefore build "local" emulators that are only valid over a relevant fraction of the input parameter space. This is achieved by running a computationally cheap full RT simulation for which case-specific computational requirements control what domain size and time period can be used, to create a first set of training points. This initial training set is then augmented by kernel density sampling to provide more input diversity while simultaneously honoring as much as possible the complex between-input dependencies. The resulting training set is then either used to train the DNN emulator or serves as training base for the kNN emulator. Our results indicate that this strategy works to some extent but does no longer deliver nearperfect predictions, as both the HPx py -DNN and HPx py -kNN outputs show discrepancies in some of the emulated time series of 2D concentration profiles compared to the baseline. Nevertheless, these deviations can presumably be smoothed out to some extent using post-filtering. Although smaller than for the 4-components system, the speedups achieved for this 7-component system remain attractive: 4.0 to 4.7 (13.4 to 16.3) and 2.4 to 2.9 (7.9 to 10.0) compared to four-threaded (single-threaded) HPx for HPx py -DNN and HPx py -kNN, respectively. Defining the maximum possible speedup as the computational gain in RT simulation that would be obtained if the geochemical calculations would come at no cost, we find that HPx py -DNN achieves a close to optimal speedup while HPx py -kNN provides half the maximum possible speedup. How to improve both accuracy and speedup for the considered cement systems and other problems will be the scope of future work.

show abstract

Chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

Abstract: The computational costs associated with coupled reactive transport simulations are mostly due to the chemical subsystem: replacing it with a pre-trained statistical surrogate is a promising strategy to achieve decisive speedups at

Cited by 2 publications

References 20 publications

DecTree v1.0 – chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

DecTree v1.0 – chemistry speedup in reactive transport simulations: purely data-driven and physics-based surrogates

Speeding up reactive transport simulations in cement systems by surrogate geochemical modeling: deep neural networks and k-nearest neighbors

Contact Info

Product

Resources

About