Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “ensemble modeling system”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 595 records · Page 33
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Protocol for applying a network-enabled gene discovery pipeline to non-model plant species

Identifying upstream regulators of key genes is essential for understanding gene regulatory mechanisms and translating these insights into functional targets. Here, we present a protocol for applying the network-enabled gene discovery pipeline (NEEDLE) to non-model plant species. We describe steps for environment setup, data preparation, computational analysis, expected outputs, and parameter considerations. NEEDLE integrates RNA sequencing (RNA-seq) processing, weighted gene co-expression analysis (WGCNA), Gene Network Inference with Ensemble of trees (GENIE3), and promoter conservation analysis to prioritize candidate transcriptional regulators.

Plant Sciences↗
DOE OSTI2023

Comparing emulation methods for a high‐resolution storm surge model

Abstract Realistic simulations of complex systems are fundamental for climate and environmental studies. Large computer systems are often not sufficient to run sophisticated computational models for large numbers of different input settings. Statistical surrogate models, or emulators, are key tools enabling fast exploration of the simulator input space. Gaussian processes have become standard for computer simulator emulation. However, they require careful implementation to scale appropriately, motivating alternative methods more recently introduced. We present a comparison study of surrogates of the Sea, Lake, and Overland Surges from Hurricanes (SLOSH) simulator—the simulator of choice for government agencies—using four emulation approaches: BASS; BART; SEPIA; and RobustGaSP. SEPIA and RobustGaSP use Gaussian processes, BASS implements adaptive splines, and BART is based on ensembles of regression trees. We describe the four models and compare them in terms of computation time and predictive metrics. These surrogates use proven and distinct methodologies, are available through accessible software, and quantify prediction uncertainty. Our data cover millions of response values. We find that SEPIA and RobustGaSP provide exceptional predictive power, but cannot scale to emulate experiments as large as the one considered in this paper as effectively as BASS and BART.

Hutchings, Grant↗
DOE OSTI2022

Towards a Unified Setup to Simulate Mid‐Latitude and Tropical Mesoscale Convective Systems at Kilometer‐Scales

Abstract Mesoscale convective systems (MCSs) are the main source of precipitation in the tropics and parts of the mid‐latitudes and are responsible for high‐impact weather worldwide. Studies showed that deficiencies in simulating mid‐latitude MCSs in state‐of‐the‐art climate models can be alleviated by kilometer‐scale models. However, whether these models can also improve tropical MCSs and whether we can find model settings that perform well in both regions is understudied. We take advantage of high‐quality MCS observations collected over the Atmospheric Radiation Measurement (ARM) facilities in the US Southern Great Plains (SGP) and the Amazon basin near Manaus (MAO) to evaluate a perturbed physics ensemble of simulated MCSs with 4 km horizontal grid spacing. A new model evaluation method is developed that enables to distinguish biases stemming from spatiotemporal displacements of MCSs from biases in their reflectivity and cloud shield. Amazon MCSs are similarly well simulated across these evaluation metrics than SGP MCSs despite the challenges anticipated from weaker large‐scale forcing in the tropics. Generally, SGP MCSs are more sensitive to the choice of model microphysics, while Amazon cases are more sensitive to the planetary boundary layer (PBL) scheme. Although our tested model physics combinations had strengths and weaknesses, combinations that performed well for SGP simulations result in worse results in the Amazon basin and vice versa. However, we identified model settings that perform well at both locations, which include the Thompson and Morrison microphysics coupled with the Yonsei University (YSU) PBL scheme and the Thompson scheme coupled with the Mellor‐Yamada‐Janjic PBL scheme.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

The overlapping fragment approach for non-orthogonal configuration interaction with fragments

The non-orthogonal configuration interaction with fragments (NOCI-F) approach is extended opening the possibility to study intramolecular processes and materials with covalent or ionic lattices. So far, NOCI-F has been applied to study intermolecular energy and electron transfer employing ensembles of fragments that do not have atoms or bonds in common. The here presented approach divides the target system into two overlapping fragments that share one or more atoms and/or one or more bonds. After the construction of a collection of (multiconfigurational) fragment wave functions in a state specific optimization procedure, the fragment wave functions are combined to form many-electron basis functions for the non-orthogonal configuration interaction of the whole system. The orbitals in the overlapping fragment are defined by a corresponding orbital transformation of the fragment orbitals through a singular value decomposition. The overlapping fragments approach is first illustrated for a model system and then used to highlight some possible applications of NOCI with overlapping fragments. In conclusion, the results of excited state diffusion in transition metal oxide, intramolecular singlet fission and magnetic interactions in organic biradicals and ionic compounds are discussed.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Models, data, and scripts associated with “Prediction of Distributed River Sediment Respiration Rates using Community-Generated Data and Machine Learning”

This data package is associated with the publication “Prediction of Distributed River Sediment Respiration Rates using Community-Generated Data and Machine Learning’’ submitted to the Journal of Geophysical Research: Machine Learning and Computation (Scheibe et al. 2024). River sediment respiration observations are expensive and labor intensive to obtain and there is no physical model for predicting this quantity. The Worldwide Hydrobiogeochemisty Observation Network for Dynamic River Systems (WHONDRS) observational data set (Goldman et al.; 2020) is used to train machine learning (ML) models to predict respiration rates at unsampled sites. This repository archives training data, ML models, predictions, and model evaluation results for the purposes of reproducibility of the results in the associated manuscript and community reuse of the ML models trained in this project. One of the key challenges in this work was to find an optimum configuration for machine learning models to work with this feature-rich (i.e. 100+ possible input variables) data set. Here, we used a two-tiered approach to managing the analysis of this complex data set: 1) a stacked ensemble of ML models that can automatically optimize hyperparameters to accelerate the process of model selection and tuning and 2) feature permutation importance to iteratively select the most important features (i.e. inputs) to the ML models. The major elements of this ML workflow are modular, portable, open, and cloud-based, thus making this implementation a potential template for other applications. This data package is associated with the GitHub repository found at Please see the file level metadata (flmd; “sl-archive-whondrs_flmd.csv”) for a list of all files contained in this data package and descriptions for each. Please see the data dictionary (dd; “sl-archive-whondrs_dd.csv”) for a list of all column headers contained within comma separated value (csv) files in this data package and descriptions for each. The GitHub repository is organized into five top-level directories: (1) “input_data” holds the training data for the ML models; (2) “ml_models” holds machine learning models trained on the data in “input_data”; (3) “scripts” contains data preprocessing and postprocessing scripts and intermediate results specific to this data set that bookend the ML workflow; (4) “examples” contains the visualization of the results in this repository including plotting scripts for the manuscript (e.g., model evaluation, FPI results) and scripts for running predictions with the ML models (i.e., reusing the trained ML models); (5) “output_data” holds the overall results of the ML model on that branch. Each trained ML model resides on its own branch in the repository; this means that inputs and outputs can be different branch-to-branch. Furthermore, depending on the number of features used to train the ML models, the preprocessing and postprocessing scripts, and their intermediate results, can also be different branch-to-branch. The “main-*” branches are meant to be starting points (i.e. trunks) for each model branch (i.e. sprouts). Please see the Branch Navigation section in the top-level README.md in the GitHub repository for more details. There is also one hidden directory “.github/workflows”. This hidden directory contains information for how to run the ML workflow as an end-to-end automated GitHub Action but it is not needed for reusing the ML models archived here. Please the top-level README.md in the GitHub repository for more details on the automation.

13C↗
DOE OSTI2021

Opacity Table Generation

Radiation opacity is a key property of material systems which describes the absorption and scattering of photons. Such processes are an important part of energy transport, diagnostics, and astrophysical phenomena. Recent experiments in iron opacity and continuum lowering have sparked renewed interest in and careful vetting of opacity models. Even within the local thermodynamic equilibrium approximation, opacity varies over a large parameter space: material, temperature, density, and energy. Opacity tables that span wide regimes of this parameter space are of prime importance to radiation-hydrodynamics codes. Accurate opacity calculations cannot be achieved inline during a radiation hydrodynamics simulation since they involve accounting for all possible electron transitions of all important atomic states of a thermal ensemble. Livermore’s current framework utilizes precalculated tables at many densities, temperatures, and energies. Last year, the opacity theory team completed an L2 milestone concerning the development of a new code, Opus, which served the dual purpose of creating a modern code infrastructure and version control as well as the basis for training new members of that team, a critical need as key members near or are passed retirement. For the current L2 milestone, we exercised and validated Opus on three elements: boron, carbon, and nitrogen, which have few enough electrons to allow a careful convergence study over many input parameters. Here, we report on the production of new tables, show their convergence with respect to important parameters, state some of their limitations, and document a mostly automated production process. In this section, we address the milestone completion criteria, found on the title page and the subsection headings. In section 2, we validate Opus against Tycho and compare with TabOp and Atomic. Section 3 details some of the physics implemented in Opus and relevant to this report. We then cover details of the density-temperature grid in section 4 and do a spectral comparison between different parameter settings and different codes in section 5. After summarizing these results in section 6, we give instructions how to run the opacity table scripts in appendix A and document the scripts and input/output files in appendix B.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Nanoparticle Tracking to Probe Transport in Porous Media

From the granular and fractured subsurface environment to highly engineered polymer membranes used in pharmaceutical purification, porous materials are ubiquitous in nature and industrial applications. In particular, porous media are used extensively in processes including water treatment, pharmaceutical sterilization, food/beverage processing and heterogeneous catalysis, where hindered mass transport is either essential to the process or a necessary, but undesirable limitation. Unfortunately, there are currently no universal models capable of predicting mass transport based on a description of the porous material because real porous materials are complex, and because many coupled dynamic mechanisms (e.g. adsorption, steric effects, hydrodynamic effects, electrostatic interactions, etc.) give rise to the observed macroscopic transport phenomena. While classical techniques, like nuclear magnetic resonance and dynamic light scattering, provide useful information about mass transport in porous media at the ensemble level, they provide limited insight into the microscopic mechanisms that give rise to complex phenomena such as anomalous diffusion, hindered pore-space accessibility, and unexpected retention under flow, among many others. To address this issue, we have developed refractive index matching imaging systems, combined with single-particle tracking methods, allowing the direct visualization of single particle motion within a variety of porous materials. Here, we summarize our recent efforts to advance the understanding of nanoparticle transport in porous media using single-particle tracking methods in both fundamental and applied scenarios. First, we describe the basic principles for two-dimensional and three-dimensional single-particle tracking in porous materials. Then, we describe concrete examples of nanoparticle transport in porous materials from two perspectives: (1) understanding fundamental elementary particle transport processes in porous media, including pore accessibility and cavity escape, which limit transport in porous media; (2) facilitating applications in industrial processes, e.g. by understanding the mechanisms of particle fouling and remobilization in filtration membranes. Finally, we provide an outlook of opportunities associated with investigating other types of mass transport in confined environments using single-particle tracking methods, including electrophoretic and self-propelled motion.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Microelectrode Arrays for Electrochemical Cycling of Individual Battery Particles

The electrochemical characterization of battery materials is usually performed on porous electrodes containing conductive additives, binders, and ensembles of redox-active particles at densities as high as 10 15 cm –3 . These complex structures often obscure individual battery particles’ intrinsic properties, hindering insights into the fundamentals of battery electrochemistry. To address this challenge, we developed microelectrode arrays to enable the charge and discharge of individual battery particles in liquid electrolytes. In this Methods and Protocols paper, we describe the fabrication of the microelectrode array chips, the assembly of particles onto microelectrodes, and the different types of electrochemical techniques that can be used, including galvanostatic cycling, cyclic voltammetry, potentiostatic intermittent titration, and electrochemical impedance spectroscopy. We use micron-sized lithium cobalt oxide as a model system to showcase the reliability and versatility of single-particle electrochemical experiments enabled by our microelectrode array. In conclusion, we aim to provide a detailed description to promote future uses of microelectrode arrays in fundamental research of batteries and other electrochemical systems.

25 ENERGY STORAGE↗
DOE OSTI2025

Global phase diagram of D-wave superconductivity in the square-lattice t-J model

The Hubbard and closely related t-J models are exciting platforms for unconventional superconductivity (SC). Through state-of-the-art density matrix renormalization group calculations using the grand canonical ensemble, we address open issues regarding the ground-state phase diagram of the extended t-J model on a square lattice in the parameter regime relevant to cuprate superconductors. On large 8-leg cylinders, we demonstrate that the pure t-J model with only nearest-neighbor hoppings and superexchange interactions, for a wide range of doping ( δ = 0.1 to 0.2 ), hosts robust d-wave superconductivity possibly coexisting with weak unidirectional pair density wave. Furthermore, a small next nearest neighbor hopping t 2 suppresses pair and charge density waves, resulting in a uniform d-wave SC phase in both electron- and hole-doped cuprate model systems. Our work validates the t-J model as a proper minimum model for the emergence of superconductivity in cuprate superconductors.

Chen, Feng (ORCID:000000018196030X)↗
DOE OSTI2022

Accurate and Rapid Forecasts for Geologic Carbon Storage via Learning-Based Inversion-Free Prediction

Carbon capture and storage (CCS) is one approach being studied by the U.S. Department of Energy to help mitigate global warming. The process involves capturing CO 2 emissions from industrial sources and permanently storing them in deep geologic formations (storage reservoirs). However, CCS projects generally target “green field sites,” where there is often little characterization data and therefore large uncertainty about the petrophysical properties and other geologic attributes of the storage reservoir. Consequently, ensemble-based approaches are often used to forecast multiple realizations prior to CO 2 injection to visualize a range of potential outcomes. In addition, monitoring data during injection operations are used to update the pre-injection forecasts and thereby improve agreement between forecasted and observed behavior. Thus, a system for generating accurate, timely forecasts of pressure buildup and CO 2 movement and distribution within the storage reservoir and for updating those forecasts via monitoring measurements becomes crucial. This study proposes a learning-based prediction method that can accurately and rapidly forecast spatial distribution of CO 2 concentration and pressure with uncertainty quantification without relying on traditional inverse modeling. The machine learning techniques include dimension reduction, multivariate data analysis, and Bayesian learning. The outcome is expected to provide CO 2 storage site operators with an effective tool for timely and informative decision making based on limited simulation and monitoring data.

58 GEOSCIENCES↗
DOE OSTI2023

Variational deep learning of equilibrium transition path ensembles

Here, we present a time-dependent variational method to learn the mechanisms of equilibrium reactive processes and efficiently evaluate their rates within a transition path ensemble. This approach builds off of the variational path sampling methodology by approximating the time-dependent commitment probability within a neural network ansatz. The reaction mechanisms inferred through this approach are elucidated by a novel decomposition of the rate in terms of the components of a stochastic path action conditioned on a transition. This decomposition affords an ability to resolve the typical contribution of each reactive mode and their couplings to the rare event. The associated rate evaluation is variational and systematically improvable through the development of a cumulant expansion. We demonstrate this method in both over- and under-damped stochastic equations of motion, in low-dimensional model systems, and in the isomerization of a solvated alanine dipeptide. In all examples, we find that we can obtain quantitatively accurate estimates of the rates of the reactive events with minimal trajectory statistics and gain unique insights into transitions through the analysis of their commitment probability.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Fundamental advantages of multijunction thermoradiative cells

Thermoradiative (TR) cells convert heat to work through emission of thermal radiation. Multijunction thermoradiative cells have received little research interest due to the apparent overlap with energy harvesting limits. Through detailed balance formalism, the present study models both single- and multi-diode TR devices, establishing their performance limits and identifying key factors that influence efficiency. For single-junction TR cells, we derive a relationship for the optimal bandgap and show that higher emitter temperatures increase efficiency. When the receiver is at absolute zero temperature, multijunction TR cells have little advantage over the single-junction cell in terms of the maximum output power density. However, for higher receiver temperatures, the multijunction configuration demonstrates significant improvements in both power and efficiency through the optimization of the chemical potential and bandgap of each diode in the complete ensemble. A 500 K emitter and 300 K receiver TR system can achieve a 21.25% increase in efficiency and a 10% improvement in the power density with multijunction architecture compared to a single junction through chemical potential optimization. These findings suggest that multijunction TR cells offer a promising approach to advancing low-grade heat recovery technologies for efficient heat-to-work conversion.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2020

Entanglement Hamiltonian of Many-Body Dynamics in Strongly Correlated Systems

A powerful perspective in understanding nonequilibrium quantum dynamics is through the time evolution of its entanglement content. Yet apart from a few guiding principles for the entanglement entropy,to date, much less is known about the refined characteristics of entanglement propagation. In this paper we unveil signatures of the entanglement evolving and information propagating out of equilibrium, from the view of the entanglement Hamiltonian. We investigate quantum quench dynamics of prototypical Bose-Hubbard model using state-of-the-art numerical technique combined with conformal field theory. Before reaching equilibrium, it is found that a current operator emerges in the entanglement Hamiltonian, implying that entanglement spreading is carried by particle flow. In the long-time limit the subsystem enters a steady phase, evidenced by the dynamic convergence of the entanglement Hamiltonian to the expectation of a thermal ensemble. Importantly, the entanglement temperature in steady state is spatially independent,which provides an intuitive trait of equilibrium. These findings not only provide crucial information on how equilibrium statistical mechanics emerges in many-body dynamics, but also add a tool to exploring quantum dynamics from the perspective of the entanglement Hamiltonian.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Sensitivity of organized convective storms to model grid spacing in current and future climates

Mesoscale convective systems (MCSs) are complexes of thunderstorms that become organized and cover hundreds of kilometres over several hours. MCSs are prolific rain producers in the tropics and mid-latitudes and are the major cause of warm-season flooding. Traditionally, climate models have difficulties in simulating MCSs partly due to the misrepresentation of complex process interactions that operate across a large range of scales. Significant improvements in simulating MCSs have been found in kilometre-scale models that explicitly simulate deep convection. However, these models operate in the grey zone of turbulent motion and have known deficiencies in simulating small-scale processes (e.g. entrainment, vertical mass transport). Here, we perform mid-latitude idealized ensemble MCS simulations under current and future climate conditions in three atmospheric regimes: hydrostatic (12 km horizontal grid spacing; Δx), non-hydrostatic (Δx = 4, 2 and 1 km) and large eddy scale (Δx = 500 m and 250 m). Our results show a dramatic improvement in simulating MCS precipitation, movement, cold pools, and cloud properties when transitioning from 12 km to 4 km Δx. Decreasing Δx beyond 4 km results in modest improvements except for up- and downdraft sizes, average vertical mass fluxes, and cloud top height and temperature, which continue to change. Most important for climate modelling is that Δx = 4 km simulations reliably capture most MCS climate change signals compared to those of the Δx = 250 m runs. Significantly different climate change signals are found in Δx = 12 km runs that overestimate extreme precipitation changes by up to 100%.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Loosely Conditioned Emulation of Global Climate Models With Generative Adversarial Networks

Climate models encapsulate our best understanding of the Earth system, allowing research to be conducted on its future under alternative assumptions of how human-driven climate forces are going to evolve. An important application of climate models is to provide metrics of mean and extreme climate changes, particularly under these alternative future scenarios, as these quantities drive the impacts of climate on society and natural systems. Because of the need to explore a wide range of alternative scenarios and other sources of uncertainties in a computationally efficient manner, climate models can only take us so far, as they require significant computational resources, especially when attempting to characterize extreme events, which are rare and thus demand long and numerous simulations in order to accurately represent their changing statistics. Here we use deep learning in a proof of concept that lays the foundation for emulating global climate model output for different scenarios. We train two "loosely conditioned" Generative Adversarial Networks (GANs) that emulate daily precipitation output from a fully coupled Earth system model: one GAN modeling Fall-Winter behavior and the other Spring-Summer. Our GANs are trained to produce spatiotemporal samples: 32 days of precipitation over a 64x128 regular grid discretizing the globe. We evaluate the generator with a set of related performance metrics based upon KL divergence, and find the generated samples to be nearly as well matched to the test data as the validation data is to test. We also find the generated samples to accurately estimate the mean number of dry days and mean longest dry spell in the 32 day samples. Our trained GANs can rapidly generate numerous realizations at a vastly reduced computational expense, compared to large ensembles of climate models, which greatly aids in estimating the statistics of extreme events.

climate emulation, extreme climate, impacts, machi↗
DOE OSTI2022

A Bayesian Deep Learning Approach to Near-Term Climate Prediction

Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state-of-the-art decadal climate prediction efforts, we pursue a complementary machine-learning-based approach to climate prediction. The example problem setting we consider consists of predicting natural variability of the North Atlantic sea surface temperature on the interannual timescale in the pre-industrial control simulation of the Community Earth System Model. While previous works have considered the use of recurrent networks such as convolutional LSTMs and reservoir computing networks in this and other similar problem settings, we currently focus on the use of feedforward convolutional networks. In particular, we find that a feedforward convolutional network with a Densenet architecture is able to outperform a convolutional LSTM in terms of predictive skill. Next, we go on to consider a probabilistic formulation of the same network based on Stein variational gradient descent and find that in addition to providing useful measures of predictive uncertainty, the probabilistic (Bayesian) version improves on its deterministic counterpart in terms of predictive skill. Finally, we characterize the reliability of the ensemble of machine learning models obtained in the probabilistic setting by using analysis tools developed in the context of ensemble numerical weather prediction.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Pooling Data Improves Multimodel IDF Estimates over Median-Based IDF Estimates: Analysis over the Susquehanna and Florida

Traditional multimodel methods for estimating future changes in precipitation intensity, duration, and frequency (IDF) curves rely on mean or median of models’ IDF estimates. Such multimodel estimates are impaired by large estimation uncertainty, shadowing their efficacy in planning efforts. Here, assuming that each climate model is one representation of the underlying data generating process, i.e., the Earth system, we propose a novel extension of current methods through pooling model data: (i) evaluate performance of climate models in simulating the spatial and temporal variability of the observed annual maximum precipitation (AMP), (ii) bias-correct and pool historical and future AMP data of reasonably performing models, and (iii) compute IDF estimates in a nonstationary framework from pooled historical and future model data. Pooling enhances fitting of the extreme value distribution to the data and assumes that data from reasonably performing models represent samples from the “true” underlying data generating distribution. Through Monte Carlo simulations with synthetic data, we show that return periods derived from pooled data have smaller biases and lesser uncertainty than those derived from ensembles of individual model data. We apply this method to NA-CORDEX models to estimate changes in 24-h precipitation intensity–frequency (PIF) estimates over the Susquehanna watershed and Florida peninsula. Our approach identifies significant future changes at more stations compared to median-based PIF estimates. The analysis suggests that almost all stations over the Susquehanna and at least two-thirds of the stations over the Florida peninsula will observe significant increases in 24-h precipitation for 2–100-yr return periods.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Deep learning to estimate permeability using geophysical data

Time-lapse electrical resistivity tomography (ERT) is a popular geophysical method to estimate three-dimensional (3D) permeability fields from electrical potential difference measurements. Traditional inversion and data assimilation methods are used to ingest this ERT data into hydrogeophysical models to estimate permeability. Due to ill-posedness and the curse of dimensionality, existing inversion strategies provide poor estimates and low resolution of the 3D permeability field. Recent advances in deep learning provide us with powerful algorithms to overcome this challenge. This paper presents a deep learning (DL) framework to estimate the 3D subsurface permeability from time-lapse ERT data. To test the feasibility of the proposed framework, we train DL-enabled inverse models on simulation data. Each measurement in both synthetic and field data is standardized by removing the mean and scaling the time-series to unit variance. This pre-processing step is necessary to bring simulation data closer to field observations. Subsurface process models based on hydrogeophysics are used to generate this synthetic data. Training performed on limited simulation data resulted in the DL model over-fitting. An advanced data augmentation based on mixup is implemented to generate additional training samples to overcome this issue. This mixup technique creates weakly labeled (low-fidelity) samples from strongly labeled (high-fidelity) data. The weakly labeled training data is then used to develop DL-enabled inverse models and reduce over-fitting. As both time-lapse ERT (1133048 features/realization) and 3D permeability (585453 features/realization) data samples are from a high-dimensional space, principal component analysis (PCA) is employed to reduce dimensionality. Encoded ERT and encoded permeability are generated using the trained PCA estimators. A deep neural network is then trained to map the encoded ERT to encoded permeability. This mixup training and unsupervised learning allowed us to build a fast and reasonably accurate DL-based inverse model under limited simulation data. Results show that proposed weak supervised learning can capture salient spatial features in the 3D permeability field. Quantitatively, the average mean squared error (in terms of the natural log) on the strongly labeled training, validation, and test datasets is less than 0.5. The R 2 -score (global metric) is greater than 0.75, and the percent error in each cell (local metric) is less than 10%. Finally, an added benefit in terms of computational cost is that the proposed DL-based inverse model is at least O(10 4 ) times faster than running a forward model once it is trained. Data generation, DL model training, and hyperparameter tuning to identify optimal neural network architectures utilized high-performance computing resources while the DL inference is performed on a standard laptop. Approximately, O(10 5 ) processor hours are used for generating data and DL tuning and training. We acknowledge that the data generation and DL model development are expensive. But once a DL model is trained, it can be re-used for inversion rapidly for the given system, with set physics and domain. Note that traditional inversion may require multiple forward model simulations (e.g., in the order of 10 to 1000), which are very expensive. This computational savings ≈ O(10 5 ) – O(10 7 )) makes the proposed DL-based inverse model attractive for subsurface imaging and real-time ERT monitoring applications due to fast and yet reasonably accurate estimations of permeability field.

58 GEOSCIENCES↗