Abstract:This paper presents a physics-informed machine learning (ML) framework to construct reduced-order models (ROMs) for reactive-transport quantities of interest (QoIs) based on high-fidelity numerical simulations. QoIs include species decay, product yield, and degree of mixing. The ROMs for QoIs are applied to quantify and understand how the chemical species evolve over time. First, high-resolution datasets for constructing ROMs are generated by solving anisotropic reaction-diffusion equations using a non-negative finite element formulation for different input parameters. Non-negative finite element formulation ensures that the species concentration is non-negative (which is needed for computing QoIs) on coarse computational grids even under high anisotropy. The reactive-mixing model input parameters are a time-scale associated with flipping of velocity, a spatial-scale controlling small/large vortex structures of velocity, a perturbation parameter of the vortex-based velocity, anisotropic dispersion strength/contrast, and molecular diffusion. Second, random forests, F-test, and mutual information criterion are used to evaluate the importance of model inputs/features with respect to QoIs. Third, Support Vector Machines (SVM) and Support Vector Regression (SVR) are used to construct ROMs based on the model inputs. Then, SVR-ROMs are used to predict scaling of QoIs. Qualitatively, SVR-ROMs are able to describe the trends observed in the scaling law associated with QoIs. Fourth, the scaling law's exponent dependence on model inputs/features are evaluated using $k$-means clustering. Finally, in terms of the computational cost, the proposed SVM-ROMs and SVR-ROMs are $\mathcal{O}(10^7)$ times faster than running a high-fidelity numerical simulation for evaluating QoIs.

What problem does this paper attempt to address?

The main problem this paper attempts to address is the development of a Physics-Informed Machine Learning (PIML) framework for constructing Reduced-Order Models (ROMs) of Quantities of Interest (QoIs) in reactive transport processes. Specifically, the paper focuses on how to efficiently and accurately predict the decay of chemical species, product formation, and the degree of mixing during reactive mixing processes. ### Main Problems 1. **Quantifying the impact of input parameters on chemical species mixing, reactant decay, and product formation**: - The paper generates datasets through high-fidelity numerical simulations and uses methods such as Random Forest, F-test, and Mutual Information criteria to assess the importance of different input parameters on QoIs. - The results indicate that anisotropic dispersion intensity/contrast is the most important factor, while the time scale associated with velocity reversal is the least important factor. 2. **Efficient and accurate prediction of reactive mixing states at different time stages**: - ROMs are constructed using Support Vector Machines (SVM) and Support Vector Regression (SVR). - SVR-ROMs can qualitatively describe the scaling law trends of QoIs, and their R² scores are greater than 0.9 when predicting the scaling relationships of QoIs, indicating good quantitative predictive capability. - Through k-means clustering analysis, it is inferred that incomplete mixing occurs under high anisotropic contrast, while the system is fully mixed under low anisotropic contrast. 3. **Improving computational efficiency**: - The proposed SVM-ROMs and SVR-ROMs are approximately 10^7 times faster than running high-fidelity numerical simulations, making these models very attractive for real-time monitoring and reactive transport sensing applications. ### Contributions - **Physics-Informed Machine Learning Models**: Developed physics-informed machine learning models for predicting the state and progress of reactive mixing. - **High-Fidelity Numerical Simulations**: Model construction is based on high-fidelity numerical simulation data that respects the underlying physics. - **Computational Efficiency and Data Compression**: The computational efficiency of the models is significantly improved, with a data compression ratio of approximately 10^-4, and with a small amount of training data (e.g., 5%), the prediction accuracy for unseen data exceeds 90%. ### Application Prospects - **Uncertainty Quantification Studies**: Due to the improved computational efficiency and accurate predictive capability, this method is suitable for comprehensive uncertainty quantification studies that require thousands of forward model runs. - **Real-Time Monitoring and Sensing Applications**: The rapid computational capability of the models makes them practically valuable for real-time monitoring and reactive transport sensing applications. In summary, this paper aims to develop efficient reduced-order models by combining physical information and machine learning techniques to accurately predict key quantities in reactive mixing processes, thereby improving computational efficiency and predictive accuracy.

Physics-Informed Machine Learning Models for Predicting the Progress of Reactive-Mixing

A Comparative Study of Machine Learning Models for Predicting the State of Reactive Mixing

Projection-based Reduced Order Modeling of Multi-Species Mixing and Combustion

Parametric reduced order models with machine learning for spatial emulation of mixing and combustion problems

Scientific Machine Learning Based Reduced-Order Models for Plasma Turbulence Simulations

Reduced-order Modeling of Solid-Liquid Mixing in a Stirred Tank Using Data-Driven Singular Value Decomposition

Proper Orthogonal Decomposition Mode Coefficient Interpolation: A Non-Intrusive Reduced-Order Model for Parametric Reactor Kinetics

Nonlinear dimensionality reduction then and now: AIMs for dissipative PDEs in the ML era

An Efficient Data-Driven Multiscale Stochastic Reduced Order Modeling Framework for Complex Systems

Traditional machine learning methods in predicting the physics of subcritical systems in source-equilibrium

Dynamic Process Optimization Using a Reduced Order Modeling Framework

Reduced order model of diffusion flames based on multi-scale data from detailed CFD: the impact of preprocessing

A Reduced Order Model conditioned on monitoring features for estimation and uncertainty quantification in engineered systems

Deep-learning-based reduced-order modeling for subsurface flow simulation

A LSTM-enhanced surrogate model to simulate the dynamics of particle-laden fluid systems

Nonintrusive projection-based reduced order modeling using stable learned differential operators

Train Small, Model Big: Scalable Physics Simulators via Reduced Order Modeling and Domain Decomposition

Reduced-Order Modeling for Multiphase Flow Using a Physics-Based Deep Learning

Scalable physics-guided data-driven component model reduction for steady Navier-Stokes flow

Parametric Intrusive Reduced Order Models enhanced with Machine Learning Correction Terms

Adaptive physics-informed neural operator for coarse-grained non-equilibrium flows