Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “data predictive control”

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 1,693 records · Page 94
All sourcesNASA NTRSDOE OSTI
NASA NTRS2011

Current-Sensitive Path Planning for an Underactuated Free-Floating Ocean Sensorweb

This work investigates multi-agent path planning in strong, dynamic currents using thousands of highly under-actuated vehicles. We address the specific task of path planning for a global network of ocean-observing floats. These submersibles are typified by the Argo global network consisting of over 3000 sensor platforms. They can control their buoyancy to float at depth for data collection or rise to the surface for satellite communications. Currently, floats drift at a constant depth regardless of the local currents. However, accurate current forecasts have become available which present the possibility of intentionally controlling floats' motion by dynamically commanding them to linger at different depths. This project explores the use of these current predictions to direct float networks to some desired final formation or position. It presents multiple algorithms for such path optimization and demonstrates their advantage over the standard approach of constant-depth drifting.

oceanography↗
NASA NTRSDate not supplied

Pressure Gradient and Gas Hold-Up in Two-Phase Flows through Porous Media in Microgravity

Single and two-phase flow through porous media are encountered on ground and in space in numerous applications related to life support systems, fuel cells, chemical/materials processing and transporting nutrients to plants. These systems operate differently in the microgravity environment encountered in space travel because the density differences no longer cause the phases to separate or “drain” under the body force of gravity. In the absence of gravity, the interfacial or capillary forces play a more significant role in determining the operational variables such as phase distribution, liquid holdup, and pressure drop especially when the liquid inertia and viscous forces are at minimum. Lower liquid flow rates in air-water two-phase flows represent the case for most of the space processes that depend on two-phase flows in porous media. The Packed Bed Reactor Experiment (PBRE) was developed and flown on the International Space Station (ISS) by NASA to extend a series of fundamental studies of gas-liquid flows through porous media. The goal of the this series of flight experiments (PBRE-1 and PBRE-2) is to better understand the hydrodynamics of two phase flows in porous media in flow regimes dominated by inertia and viscous effects and understanding the role of the interfacial forces and their role in determining the flow dynamics. The flight experiments were preceded by a series of low gravity experiments performed on the low gravity aircraft. These experiments led to the development of a semi-empirical two-phase pressure drop and flow pattern transition in support of a number of reactor beds planned for water reclamation processes onboard of the international space station. The PBRE flight experiment was designed to deliver a wide range of tightly controlled gas (nitrogen) and liquid (water) flows to one of two interchangeable test sections differing only in the type of internal packing material. In this presentation, we describe the PBRE-2 experiment and present preliminary low gravity data on pressure drop and gas hold-up and compare with semi-empirical model pressure drop predictions.

Brian Motil↗
DOE OSTI2020

On the Formalization of Development and Assessment Process for Digital Twins in the Nearly Autonomous Management and Control System

In recent years, the autonomous control system has been encouraged in advanced reactors for restoring economic viability, simplifying the operation and maintenance, and enabling remote-site power generations [1]. Since the reactor is expected to be operated for a long period of time with a limited number of individuals onsite, it is recommended that the autonomous control system should have access to very realistic models of the state of processes in the whole lifecycle, together with these process behaviors in interaction with their environment in the real world. As a result, digital twin (DT) technology is suggested in autonomous control systems. DT is defined as a digital representation of a physical object or system, which contains a record for the histories of loads, operation and maintenance status, predictions for the near-term transient of important state variables, and decision-making process [2]. Since machine learning (ML) can recognize patterns within a complex system in real-time applications, it has been used to build DTs in the autonomous control systems for advanced reactors. Meanwhile, due to the rareness of operation data in accident scenarios, the development and assessment of DTs is expected to be mainly driven by simulations. Although the capability and feasibility of ML-based DTs are recognized in improving the safety and efficiency of reactor control, a major concern from the regulatory commission and the nuclear industry is whether the information from a DT is developed and assessed in accordance with expectation and requirements by the target decision. Such concerns not only affect the acceptance criteria for DTs, but also values that can be extracted from DTs and autonomous control system during operations. Inspired by the success of formal methods in improving the reliability and robustness of computer programming and software development, it is suggested that the development and assessment process (DAP) for both separate DTs and integral control system should be formalized in a transparent, consistent, and improvable manner. In this study, a digital-twin development and assessment process (DT-DAP) is proposed by adapting the evaluation model development and assessment process (EMDAP) [3] to requirements by the autonomous control system, ML algorithms, and DT technology. To demonstrate the framework, a baseline nearly autonomous management and control (NAMAC) system with ML-based DTs for diagnosis and prognosis is developed and assessed based on the framework. It is found that with selected testing methods and techniques, the DT-DAP can help identify errors in DTs and NAMAC which would otherwise be left unverified. Meanwhile, it is found that the DT-DAP can improve the DTs and NAMAC by continuously learning and iterating through different elements.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2022

Chromatin structure predicts survival in glioma patients

The pathological changes in epigenetics and gene regulation that accompany the progression of low-grade to high-grade gliomas are under-studied. The authors use a large set of paired atac-seq and RNA-seq data from surgically resected glioma specimens to infer gene regulatory relationships in glioma. Thirty-eight glioma patient samples underwent atac-seq sequencing and 16 samples underwent additional RNA-seq analysis. Using an atac-seq/RNA-seq correlation matrix, atac-seq peaks were paired with genes based on high correlation values (|r 2 | > 0.6). Samples clustered by IDH1 status but not by grade. Surprisingly there was a trend for IDH1 mutant samples to have more peaks. The majority of peaks are positively correlated with survival and positively correlated with gene expression. Constructing a model of the top six atac-seq peaks created a highly accurate survival prediction model (r 2 = 0.68). Four of these peaks were still significant after controlling for age, grade, pathology, IDH1 status and gender. Grade II, III, and IV (primary) samples have similar transcription factors and gene modules. However, grade IV (recurrent) samples have strikingly few peaks. Patient-derived glioma cultures showed decreased peak counts following radiation indicating that this may be radiation-induced. This study supports the notion that IDH1 mutant and IDH1 wildtype gliomas have different epigenetic landscapes and that accessible chromatin sites mapped by atac-seq peaks tend to be positively correlated with expression. The data in this study leads to a new model of treatment response wherein glioma cells respond to radiation therapy by closing open regions of DNA.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

From Data to Insights: A Covariate Analysis of the IARPA BRIAR Dataset for Multimodal Biometric Recognition Algorithms at Altitude and Range

This paper examines covariate effects on fused whole body biometrics performance in the IARPA BRIAR dataset, specifically focusing on UAV platforms, elevated positions, and distances up to 1000 meters. The dataset includes outdoor videos compared with indoor images and controlled gait recordings. Normalized raw fusion scores relate directly to predicted false accept rates (FAR), offering an intuitive means for interpreting model results. A linear model is developed to predict biometric algorithm scores, analyzing their performance to identify the most influential covariates on accuracy at altitude and range. Weather factors like temperature, wind speed, solar loading, and turbulence are also investigated in this analysis. The study found that resolution and camera distance best predicted accuracy and findings can guide future research and development efforts in long-range/elevated/UAV biometrics and support the creation of more reliable and robust systems for national security and other critical domains.

Bolme, David↗
NASA NTRS1975

Automatic documentation system extension to multi-manufacturers' computers and to measure, improve, and predict software reliability. Appendix A and B

A manual which explains how to use the DOMONIC command language is presented. The manual consists of the following sections: 1. Introduction 2. What You Must Know to Use DOMONIC 3. Functions of DOMONIC 4. Entering and Manipulating Data 5. Templates and Data Definitions 6. Recipes and Document Generation 7. Initiating A Project and 8. Entering and Changing Security Controls. The first four sections must be known by all DOMONIC users. Sections five, seven and eight describe functions normally used only by project management. Section six tells how to generate documents. The manual specifies what commands to use in performing each of the functions mentioned above.

Simmons, D. B.↗
NASA NTRS1995

Flow structure generated by perpendicular blade vortex interaction and implications for helicopter noise predictions

This report summarizes accomplishments and progress for the period ending April 1995. Much of the work during this period has concentrated on preparation for an analysis of data produced by an extensive wind tunnel test. Time has also been spent further developing an empirical theory to account for the effects of blade-vortex interaction upon the circulation distribution of the vortex and on preliminary measurements aimed at controlling the vortex core size.

Devenport, William J.↗
DOE OSTI2025

TRANSP integrated modeling code for interpretive and predictive analysis of tokamak plasmas

This paper provides a comprehensive review of the TRANSP code, a sophisticated tool for interpretive and predictive analysis of tokamak plasmas, detailing its major capabilities and features. It describes the equations for particle, power, and momentum balance analysis, as well as the poloidal field diffusion equations. The paper outlines the spatial and time grids used in TRANSP and details the equilibrium assumptions and solvers. Various models for heating and current drive and radiation, including updates to the NUBEAM model, are discussed. The handling of large-scale events such as sawtooth crashes and pellet injections is examined, along with the predictive capabilities for advancing plasma profiles. The integration of TRANSP with the ITER Integrated Modeling and Analysis Suite (IMAS) is highlighted, demonstrating enhanced data access and analysis capabilities. Additionally, the paper discusses best practices and continuous integration techniques to enhance TRANSP's robustness. The suite of TRANSP tools, designed for efficient data analysis and simulation, further supports the optimization of tokamak operations and coupling with other tokamak codes. Continuous development and support ensure that TRANSP remains a major code for the analysis of experimental data for controlled thermonuclear fusion, maintaining its critical role in supporting the optimization of tokamak operations and advancing fusion research.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS2000

Bithermal Low-Cycle Fatigue Evaluation of Automotive Exhaust System Alloy SS409

This investigation provides, for the first time, cyclic strainrange-controlled, thermomechanical fatigue results for the ferritic stainless steel alloy SS409. The alloy has seen extensive application for automotive exhaust system components. The data were generated to calibrate the Total Strain Version of the Strainrange Partitioning (TS-SRP) method for eventual application to the design and durability assessment of automotive exhaust systems. The thermomechanical cyclic lifetime and cyclic stress-strain constitutive behavior for alloy SS409 were measured using bithermal tests cycling between isothermal extremes of 400 and 800 C. Lives ranged up to 10,000 cycles to failure with hold-times of 0.33 to 2.0 minutes. The bithermal fatigue behavior is compared to isothermal, strain-controlled fatigue behavior at both 400 and 800 C. Thermomechanical cycling was found to have a profound detrimental influence on the fatigue failure resistance of SS409 compared to isothermal cycling. Supplementary bithermal cyclic stress-strain constitutive tests with hold-times ranging from 40 seconds up to 1.5 hours were conducted to calibrate the TS-SRP equation for extrapolation to longer lifetime predictions. Observed thermomechanical (bithermal) fatigue lives correlated well with lives calculated using the calibrated TS-SRP equations: 70% of the bithermal fatigue data fall within a factor of 1.2 of calculated life; 85% within a factor of 1.4; and 100% within a factor of 1.8.

Lu, Gui-Ying↗
NASA NTRS2010

Use of Quality Controlled AIRS Temperature Soundings to Improve Forecast Skill

AIRS was launched on EOS Aqua on May 4, 2002, together with AMSU-A and HSB, to form a next generation polar orbiting infrared and microwave atmospheric sounding system. The primary products of AIRS/AMSU-A are twice daily global fields of atmospheric temperature-humidity profiles, ozone profiles, sea/land surface skin temperature, and cloud related parameters including OLR. Also included are the clear column radiances used to derive these products which are representative of the radiances AIRS would have seen if there were no clouds in the field of view. All products also have error estimates. The sounding goals of AIRS are to produce 1 km tropospheric layer mean temperatures with an rms error of 1K, and layer precipitable water with an rms error of 20 percent, in cases with up to 90 percent effective cloud cover. The products are designed for data assimilation purposes for the improvement of numerical weather prediction, as well as for the study of climate and meteorological processes. With regard to data assimilation, one can use either the products themselves or the clear column radiances from which the products were derived. The AIRS Version 5 retrieval algorithm is now being used operationally at the Goddard DISC in the routine generation of geophysical parameters derived from AIRS/AMSU data. A major innovation in Version 5 is the ability to generate case-by-case level-by-level error estimates for retrieved quantities and clear column radiances, and the use of these error estimates for Quality Control. The temperature profile error estimates are used to determine a case-by-case characteristic pressure pbest, down to which the profile is considered acceptable for data assimilation purposes. The characteristic pressure p(sub best) is determined by comparing the case dependent error estimate (delta)T(p) to the threshold values (Delta)T(p). The AIRS Version 5 data set provides error estimates of T(p) at all levels, and also profile dependent values of pbest based on use of a Standard profile dependent threshold (Delta)T(p). These Standard thresholds were designed as a compromise between optimal use for data assimilation purposes, which requires highest accuracy (tighter Quality Control), and climate purposes, which requires more spatial coverage (looser Quality Control). Subsequent research using Version 5 sounding and error estimates showed that tighter Quality Control performs better for data assimilation proposes, while looser Quality Control better spatial coverage) performs better for climate purposes. We conducted a number of data assimilation experiments using the NASA GEOS-5 Data Assimilation System as a step toward finding an optimum balance of spatial coverage and sounding accuracy with regard to improving forecast skill. The model was run at a horizontal resolution of 0.5 degree latitude x 0.67 degree longitude with 72 vertical levels. These experiments were run during four different seasons, each using a different year. The AIRS temperature profiles were presented to the GEOS-5 analysis as rawinsonde profiles, and the profile error estimates (delta)T(p) were used as the uncertainty for each measurement in the data assimilation process.

Susskind, Joel↗
DOE OSTI2025

Control of machining-induced residual stress via tool geometry and process parameter modification

Distortion generated in machined, monolithic, thin-walled aerospace components due to residual stresses leads to significant material and economic waste in the manufacturing industry. Inherent residual stress (IRS) present in stock materials combines with machining-induced residual stress (MIRS) to influence the final machined part distortion. It is hypothesized that MIRS can be controlled, based on the part geometry, through deliberate cutting tool geometry and process parameter modifications to negate the effect of IRS on distortion, consequently resulting in distortion-free parts. A finite element (FE) orthogonal cutting model is developed to study how tool geometry and process parameters influence machining-induced residual stress (MIRS). Orthogonal cutting experiments are performed on Al 7075-T651 samples to measure cutting forces and MIRS. A cutting force dynamometer is used to measure forces during cutting and a novel digital image correlation (DIC) based hole drilling technique is employed to measure the near-surface residual stress (RS) in the cut samples. These data are subsequently utilized to validate the FE prediction model. Various cases of cutting simulations involving different depths of cut, tool tip radii, and rake angles are performed to study their effects on RS. Similar to prior literature, increasing the depth of cut, tool tip radius, or rake angle is found to promote the formation of near-surface tensile stresses. The competing effects of material plowing and temperature are shown to determine the type of RS at the end of the cut. Moreover, a window of variation of RS (up to ± 400 MPa) is estimated within the given range of conditions, allowing for the control of MIRS through tool and process modification.

Mathews, Ritin [ORNL] (ORCID:0000000301440828)↗
DOE OSTI2020

Long term predictions of coal fired power plant data using evolved recurrent neural networks

Abstract This work presents an investigation into the ability of recurrent neural networks (RNNs) to provide long term predictions of time series data generated by coal fired power plants. While there are numerous studies which have used artificial neural networks (ANNs) to predict coal plant parameters, to the authors’ knowledge these have almost entirely been restricted to predicting values at the next time step, and not farther into the future. Using a novel neuro-evolution strategy called Evolutionary eXploration of Augmenting Memory Models (EXAMM), we evolved RNNs with advanced memory cells to predict per-minute plant parameters and per-hour boiler parameters up to 8 hours into the future. These data sets were challenging prediction tasks as they involve spiking behavior in the parameters being predicted. While the evolved RNNs were able to successfully predict the spikes in the hourly data they did not perform very well in accurately predicting their severity. The per-minute data proved even more challenging as medium range predictions miscalculated the beginning and ending of spikes, and longer range predictions reverted to long term trends and ignored the spikes entirely. We hope this initial study will motivate further study into this highly challenging prediction problem. The use of fuel properties data generated by a new Coal Tracker Optimization (CTO) program was also investigated and this work shows that their use improved predictive ability of the evolved RNNs.

Automation & Control Systems↗
NASA NTRS1992

Code validation for the simulation of supersonic viscous flow about the F-16XL

The viewgraphs and discussion on code validation for the simulation of supersonic viscous flow about the F-16XL are provided. Because of the large potential gains related to laminar flow on the swept wings of supersonic aircraft, interest in the applications of laminar flow control (LFC) techniques in the supersonic regime has increased. A supersonic laminar flow control (SLFC) technology program is currently underway within NASA. The objective of this program is to develop the data base and design methods that are critical to the development of laminar flow control technology for application to supersonic transport aircraft design. Towards this end, the program integrates computational investigations underway at NASA Ames-Moffett and NASA Langley with flight-test investigations being conducted on the F-16XL at the NASA Ames-Dryden Research Facility in cooperation with Rockwell International. The computational goal at NASA Ames-Moffett is to integrate a thin-layer Reynolds averaged Navier-Stokes flow solver with a stability analysis code. The flow solver would provide boundary layer profiles to the stability analysis code which in turn would predict transition on the F-16XL wing. To utilize the stability analysis codes, reliable boundary layer data is necessary at off-design cases. Previously, much of the prediction of boundary layer transition has been accomplished through the coupling of boundary layer codes with stability theory. However, boundary layer codes may have difficulties at high Reynolds numbers, of the order of 100 million, and with the current complex geometry in question. Therefore, a reliable code which solves the thin-layer Reynolds averaged Navier-Stokes equations is needed. Two objectives are discussed, the first in greater depth. The first objective is method verification, via comparisons of computations with experiment, of the reliability and robustness of the code. To successfully implement LFC techniques to the F-16XL wing, the flow about the leading edge must be maintained as laminar flow. Therefore, the second objective is to focus on a series of numerical simulations with different values of angle of attack, alpha, and Reynolds numbers. The purpose of the simulations is to study their effects on the two main factors which precipitate transition to turbulence at leading edges of highly swept wings (e.g., 'spanwise contamination' and 'crossflow instability').

Flores, Jolen↗
DOE OSTI2026

Towards fully predictive gyrokinetic full- f simulations: validation and triangularity studies in TCV

Designing economical magnetic confinement fusion power plants motivates computational tools that can estimate plasma behavior from engineering parameters without direct reliance on experimental measurement of the plasma profiles. In this work, we present full-f global long-wavelength gyrokinetic simulations of edge and scrape-off layer turbulence in tokamaks that use only magnetic geometry, heating power, and particle inventory as inputs. Unlike many modeling approaches that employ free parameters fitted to experimental data, raising uncertainties when extrapolating to reactor scales. This approach directly simulates turbulence and resulting profiles through gyrokinetics without such empirical adjustments. This is achieved via an adaptive sourcing algorithm in Gkeyll that strictly controls energy injection and emulates particle sourcing due to neutral recycling. We show that the simulated kinetic profiles compare reasonably well with Thomson scattering and Langmuir probe data for Tokamak á Configuration Variable (TCV) discharge #65125, and that the simulations reproduce characteristic features such as blob transport and self-organized electric fields. Applying the same framework to study triangularity effects suggests mechanisms contributing to the improved confinement reported for negative triangularity (NT). Simulations of TCV discharges #65125 and #65130 indicate that NT increases the E x B flow shear (by about 20% in these cases), which correlates with reduced turbulent losses and a modest change in the distribution of power exhaust to the vessel wall. While the physical models contain approximations that can be refined in future work, the predictive capability demonstrated here, evolving multiple profile relaxation times with kinetic electron and ion models in hundreds of GPU hours, indicates the feasibility of using Gkeyll to support design studies of fusion devices.

Hoffmann, Antoine Cyril David [Princeton Plasma Ph↗
DOE OSTI2023

Characterizing the flux effect on the irradiation embrittlement of reactor pressure vessel steels using machine learning

In-service exposure to high-energy neutrons embrittles reactor pressure vessel (RPV) steels. An increase in the yield stress (Δσ y ) results in a corresponding increase in the brittle to ductile transition temperature (ΔT c ). Most existing models underpredict ΔT c at higher fluence following accelerated irradiations in test reactors. High fluence, up to 10 20 n/cm 2 in some cases, will be reached over extended RPV vessel operation of 80 years, or more, at low service flux. Embrittlement has been extensively studied in accelerated, higher flux test reaction irradiations. However, the use of test reactor data naturally raises the question of flux effects. Here, this study used a machine learning approach trained on a set of hardening data, covering a wide range of flux, fluence, and steel compositions to determine the interactive effects of both irradiation and material variables on Δσ y . The analysis included machine learning-based cross-plots of the variable dependence of Δσ y for six core steels (i.e., CM6, LC, LD, LG, LH, and LI), with controlled differences in their Cu and Ni contents. A primary objective is to evaluate an effective to actual fluence (Φt e /Φt) ratio, as a function of flux, fluence, and steel composition. This is information critical to properly use intermediate flux-high fluence data in calibrating a low flux-high fluence embrittlement model. The predicted Φt e /Φt is reasonably consistent with estimates previously derived from a physics-based solute recombination trap model.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Physics-informed machine learning exploration of Na storage mechanisms in disordered carbon

Sodium-ion batteries are a cost-effective, sustainable alternative to lithium-ion systems for large-scale energy storage. However, optimizing sodium storage in carbon-based anodes with microstructural complexity and atomic disorder remains a major challenge. The intrinsic inhomogeneity of these materials produces diverse local environments, making it difficult for conventional methods to predict and control ion dynamics. Hard carbon (HC) anodes, composed of ranges of ordered-to-disordered graphitic and amorphous nanodomains, offer tunable ion storage and rate capacity, yet rationale design remains a challenge due to poorly understood correlation between local atomic feature and ion transport mechanism. Here, to address this challenge, we introduce a data-driven framework that integrates validated machine-learned interatomic potentials, large-scale molecular dynamics simulations, and machine learning to elucidate sodium transport mechanisms as a function of carbon and sodium loading densities. By computing per-ion structural descriptors and applying unsupervised learning, we identify distinct diffusion modes governed by microscopic features. Supervised analysis and correlation mapping then establish quantitative links between these transport regimes and processing variables such as bulk carbon density and sodium content. This physics-informed approach establishes quantitative structure–transport relationships and offers actionable design principles for engineering high-performance HC anodes.

Data-driven framework↗
DOE OSTI2022

Octane Requirements of Lean Mixed-Mode Combustion in a Direct-Injection Spark-Ignition Engine

Here, this study investigates the octane requirements of a hybrid flame propagation and controlled autoignition mode referred to as mixed-mode combustion (MMC), which allows for strong control over combustion parameters via a spark-initiated deflagration phase. Due to the throughput limitations associated with both experiments and 3-D computational fluid dynamics calculations, a hybrid 0-D and 1-D modeling methodology was developed, supported by experimental validation data. This modeling approach relied on 1-D, two-zone engine simulations to predict bulk in-cylinder thermodynamic conditions over a range of engine speeds, compression ratios, intake pressures, trapped residual levels, fueling rates, and spark timings. Those predictions were then transferred to a 0-D chemical kinetic model, which was used to evaluate the autoignition behavior of fuels when subjected to temperature–pressure trajectories of interest. Finally, the predicted autoignition phasings were screened relative to the progress of the modeled deflagration-based combustion in order to determine if an operating condition was feasible or infeasible due to knock or stability limits. The combined modeling and experimental results reveal that MMC has an octane requirement similar to modern stoichiometric spark-ignition engines in that fuels with high research octane number (RON) and high octane sensitivity (S) enable higher loads. Experimental trends with varying RON and S were well predicted by the model for 1000 and 1400 rpm, confirming its utility in identifying the compatibility of a fuel’s autoignition behavior with an engine configuration and operating strategy. However, the model was not effective in predicting (nor designed to predict) operability limits due to cycle-to-cycle variations, which experimentally inhibited operation of some fuels at 2000 rpm. Putting the operable limits and efficiency from MMC in the context of a state-of-the-art engine, the MMC showed superior efficiencies over the range investigated, demonstrating the potential to further improve fuel economy.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2021

Data assimilation empowered neural network parametrizations for subgrid processes in geophysical flows

In the past couple of years, there has been a proliferation in the use of machine learning approaches to represent subgrid-scale processes in geophysical flows with an aim to improve the forecasting capability and to accelerate numerical simulations of these flows. Despite its success for different types of flow, the online deployment of a data-driven closure model can cause instabilities and biases in modeling the overall effect of subgrid-scale processes, which in turn leads to inaccurate prediction. To tackle this issue, we exploit the data assimilation technique to correct the physics-based model coupled with the neural network as a surrogate for unresolved flow dynamics in multiscale systems. In particular, we use a set of neural network architectures to learn the correlation between resolved flow variables and the parametrizations of unresolved flow dynamics and formulate a data assimilation approach to correct the hybrid model during their online deployment. We illustrate our framework in a set of applications of the multiscale Lorenz 96 system for which the parametrization model for unresolved scales is exactly known, and the two-dimensional Kraichnan turbulence system for which the parametrization model for unresolved scales is not known a priori. Our analysis, therefore, comprises a predictive dynamical core empowered by (i) a data-driven closure model for subgrid-scale processes, (ii) a data assimilation approach for forecast error correction, and (iii) both data-driven closure and data assimilation procedures. We show significant improvement in the long-term prediction of the underlying chaotic dynamics with our framework compared to using only neural network parametrizations for future prediction. Moreover, we demonstrate that these data-driven parametrization models can handle the non-Gaussian statistics of subgrid-scale processes, and effectively improve the accuracy of outer data assimilation workflow loops in a modular nonintrusive way.

42 ENGINEERING↗