Search NASA⌕ Search

SEARCH · Search NASA

Results for “Adaptive tuning”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Machine learning for arbitrary single-qubit rotations on an embedded device

Here, in this study, we present a technique for using machine learning (ML) for single-qubit gate synthesis on field-programmable logic for a superconducting transmon-based quantum computer based on simulated studies. Our approach is multi-stage. We first “bootstrap” a model based on simulation with access to the full state vector for measuring gate fidelity. We next present an algorithm, named adapted randomized benchmarking (ARB), for fine-tuning the gate on hardware based on measurements of the devices. We also present techniques for deploying the model on programmable devices with care to reduce the required resources. While the techniques here are applied to a transmon-based computer, many of them are portable to other architectures.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

A knowledge-informed large language model framework for U.S. nuclear power plant shutdown initiating event classification for probabilistic risk assessment

Identifying and classifying shutdown initiating events (SDIEs) is critical for developing shutdown probabilistic risk assessment for nuclear power plants. Existing computational approaches cannot achieve satisfactory performance due to the challenges of unavailable large, labeled datasets, imbalanced event types, and label noise. To address these challenges, we propose a hybrid pipeline that integrates a knowledge-informed machine learning model to prescreen non-SDIEs and a large language model (LLM) to classify SDIEs into four types. In the prescreening stage, we proposed a set of 44 SDIE text patterns that consist of the most salient keywords and phrases from six SDIE types. Text vectorization based on the SDIE patterns generates feature vectors that are highly separable by using a simple binary classifier. The second stage builds Bidirectional Encoder Representations from Transformers (BERT)-based LLM, which learns generic English language representations from self-supervised pretraining on a large dataset and adapts to SDIE classification by fine-tuning it on an SDIE dataset. The proposed approaches are evaluated on a dataset with 10,928 events using precision, recall ratio, F 1 score, and average accuracy. In conclusion, the results demonstrate that the prescreening stage can exclude more than 97% non-SDIEs, and the LLM achieves an average accuracy of 95.1% for SDIE classification.

99 - GENERAL AND MISCELLANEOUS↗
NASA NTRS1981

Tuned feedback damping with application to the Galileo spacecraft

An adaptive controller has been developed which uses phase-locked-loops as the adaptive devices. The controller provides active damping to systems which have their underdamped frequencies known only to within a particular range (e.g. plus or minus 20% uncertainty). The design and analysis of the controller is presented along with an application of such a controller to the scan platform pointing system of the Galileo spacecraft, a system with sensors and actuators separated by flexible structure.

Macala, G. A.↗
NASA NTRS1981

Considerations of open-loop, closed-loop, and adaptive multicyclic control systems

Four different types of self-tuning regulators were studied for multicyclic control of helicopter vibration. A numerical simulation of the helicopter is made, using a multivariable frequency-domain model, in terms of transfer function with six input control harmonics and six output harmonics. The model characteristics vary with flight speed. An off-line identification of model characteristics is made, using the least-squared-error method and using a succession of input and output measurements. The on-line identification of model characteristics is made using the Kalman filter solution. The optimal controls are calculated from the minimization of quadratic performance function based on response and multicyclic inputs. The performance of various regulators or controllers is judged from the stability, transient response, convergence time, and amplitude of the steady state.

Chopra, I.↗
DOE OSTI2021

Two-DoF Model-Informed Controller Gain Tuning for Several Floating Wind Platforms

Much effort has recently been put into the development of collective blade pitch controllers for floating offshore wind turbines, with the aim of overcoming negative damping issues that arise with traditional control methods. One proposed approach to this challenge involves using a two-degree-of-freedom model to inform the gain schedule of a nacelle velocity feedback term in an otherwise conventional proportional-integral controller. The model uses tower-top fore-aft and rotor angular displacements, and is used to calculate a nacelle velocity feedback gain that results in a specified increase in platform pitch damping. Earlier performance evaluations of this tuning method were favorable, suggesting its potential as an easy way for researchers to obtain an adequate controller. This paper expands on those previous results by examining the performance of the tuning method relative to baseline controllers for several hull configurations, and for several prescribed increases in platform pitch damping. Simulations were run in OpenFAST for several load cases above rated wind speed and show results consistent with trends in the earlier study. The tuning method is thus shown to be adaptable to many different types of hulls, making it useful for the evaluation of prototype designs.

artificial intelligence↗
NASA NTRS1983

Parallel architectures for iterative methods on adaptive, block structured grids

A parallel computer architecture well suited to the solution of partial differential equations in complicated geometries is proposed. Algorithms for partial differential equations contain a great deal of parallelism. But this parallelism can be difficult to exploit, particularly on complex problems. One approach to extraction of this parallelism is the use of special purpose architectures tuned to a given problem class. The architecture proposed here is tuned to boundary value problems on complex domains. An adaptive elliptic algorithm which maps effectively onto the proposed architecture is considered in detail. Two levels of parallelism are exploited by the proposed architecture. First, by making use of the freedom one has in grid generation, one can construct grids which are locally regular, permitting a one to one mapping of grids to systolic style processor arrays, at least over small regions. All local parallelism can be extracted by this approach. Second, though there may be a regular global structure to the grids constructed, there will be parallelism at this level. One approach to finding and exploiting this parallelism is to use an architecture having a number of processor clusters connected by a switching network. The use of such a network creates a highly flexible architecture which automatically configures to the problem being solved.

Gannon, D.↗
NASA NTRS2004

Smart Structures for Vibration Control on Long-Term Space Exploration and Habitation Missions

The current vision for space exploration focuses on human missions to the Moon, Mars, and beyond. To support these goals, it is certain that new vehicles and intermediate bases will be developed, whether that means simply re-direction of the ISS as a "mission research facility" or construction of a lunar base. Since these facilities are expected to be constructed from inherently light-weight materials, this work examines some of the potential sources of vibration and noise as well as means for controlling these vibrations. Many of the operating components within these facilities, such as pumps, fans, and motors, will produce vibrations during operation. These vibrations become structure in which they are housed. Resonances can impact acoustic noise levels and noise quality within the environment, possibly affecting crew health and productivity. For long-term missions in particular, it is expected that crew members will spend significant portions of their time restrained in the structure, such as in seats. As a result, the general health and well-being of the crew can be improved by limiting the harmful effects of human exposure to long-term audible and tactile vibration input. Besides the human factor, this work also examines some operational considerations in which vibrations play an important role. Vibrations can impact the environment for science and in-situ manufacturing research within these vehicles. Since a benign vibratory environment is beneficial for most types of science experiments, there is a need for various forms of vibration control. Because the operational characteristics of a vehicle can change during a long-term mission, it is further expected that the characteristics of many of the vibratory excitations will change with time. Consequently, the form of vibration control needed to improve overall habitability and usefulness of the vehicle or element for exploration missions will rely to some degree on the vibration control system's ability to adapt. To address these needs, this work also examines the development and use of smart materials to tune the dynamic characteristics of the structure in a passive sense. One prime example is the use of an adaptive electrical shunt connected to a piezoelectric patch in order to provide tuned passive vibration damping. The work also examines the use of active vibration control, such as by applying power to that same piezoelectric patch. The overall goal is to examine the use of smart structures that can react to the environment thereby improving the overall living, working, and learning environment for these long-term missions.

Gattis, Christy B.↗
NASA NTRS2014

Adaptive Augmenting Control Flight Characterization Experiment on an F/A-18

This paper summarizes the Adaptive Augmenting Control (AAC) flight characterization experiments performed using an F/A-18 (TN 853). AAC was designed and developed specifically for launch vehicles, and is currently part of the baseline autopilot design for NASA's Space Launch System (SLS). The scope covered here includes a brief overview of the algorithm (covered in more detail elsewhere), motivation and benefits of flight testing, top-level SLS flight test objectives, applicability of the F/A-18 as a platform for testing a launch vehicle control design, test cases designed to fully vet the AAC algorithm, flight test results, and conclusions regarding the functionality of AAC. The AAC algorithm developed at Marshall Space Flight Center is a forward loop gain multiplicative adaptive algorithm that modifies the total attitude control system gain in response to sensed model errors or undesirable parasitic mode resonances. The AAC algorithm provides the capability to improve or decrease performance by balancing attitude tracking with the mitigation of parasitic dynamics, such as control-structure interaction or servo-actuator limit cycles. In the case of the latter, if unmodeled or mismodeled parasitic dynamics are present that would otherwise result in a closed-loop instability or near instability, the adaptive controller decreases the total loop gain to reduce the interaction between these dynamics and the controller. This is in contrast to traditional adaptive control logic, which focuses on improving performance by increasing gain. The computationally simple AAC attitude control algorithm has stability properties that are reconcilable in the context of classical frequency-domain criteria (i.e., gain and phase margin). The algorithm assumes that the baseline attitude control design is well-tuned for a nominal trajectory and is designed to adapt only when necessary. Furthermore, the adaptation is attracted to the nominal design and adapts only on an as-needed basis (see Figure 1). The MSFC algorithm design was formulated during the Constellation Program and reached a high maturity level during SLS through simulation-based development and internal and external analytical review. The AAC algorithm design has three summary-level objectives: (1) "Do no harm;" return to baseline control design when not needed, (2) Increase performance; respond to error in ability of vehicle to track command, and (3) Regain stability; respond to undesirable control-structure interaction or other parasitic dynamics. AAC has been successfully implemented as part of the Space Launch System baseline design, including extensive testing in high-fidelity 6-DOF simulations the details of which are described in [1]. The Dryden Flight Research Center's F/A-18 Full-Scale Advanced Systems Testbed (FAST) platform is used to conduct an algorithm flight characterization experiment intended to fully vet the aforementioned design objectives. FAST was specifically designed with this type of test program in mind. The onboard flight control system has full-authority experiment control of ten aerodynamic effectors and two throttles. It has production and research sensor inputs and pilot engage/disengage and real-time configuration of up to eight different experiments on a single flight. It has failure detection and automatic reversion to fail-safe mode. The F/A-18 aircraft has an experiment envelope cleared for full-authority control and maneuvering and exhibits characteristics for robust recovery from unusual attitudes and configurations aided by the presence of a qualified test pilot. The F/A-18 aircraft has relatively high mass and inertia with exceptional performance; the F/A-18 also has a large thrust-to-weight ratio, owing to its military heritage. This enables the simulation of a portion of the ascent trajectory with a high degree of dynamic similarity to a launch vehicle, and the research flight control system can simulate unstable longitudinal dynamics. Parasitic dynamics such as slosh and bending modes, as well as atmospheric disturbances, are being produced by the airframe via modification of bending filters and the use of secondary control surfaces, including leading and trailing edge flaps, symmetric ailerons, and symmetric rudders. The platform also has the ability to inject signals in flight to simulate structural mode resonances or other challenging dynamics. This platform also offers more test maneuvers and longer maneuver times than a single rocket or missile test, which provides ample opportunity to fully and repeatedly exercise all aspects of the algorithm. Prior to testing on an F/A-18, AAC was the only component of the SLS autopilot design that had not been flight tested. The testing described in this paper raises the Technology Readiness Level (TRL) early in the SLS Program and is able to demonstrate its capabilities and robustness in a flight environment.

VanZwieten, Tannen S.↗
DOE OSTI2020

Inhibition of DNA methylation in P. soloecismus alters algae productivity

Eukaryotic organisms regulate the organization, structure, and accessibility of their genomes through chromatin remodeling that can be inherited as epigenetic modifications. These DNA and histone protein modifications are ultimately responsible for an organism’s molecular adaptation to the environment, resulting in distinctive phenotypes. Epigenetic manipulation of algae holds yet untapped potential for the optimization of biofuel production and bioproduct formation; however, epigenetic machinery and modes-of-action have not been well characterized in algae. We sought to determine the extent to which the biofuel platform species Picochlorum soloecismus utilizes DNA methylation to regulate its genome. We found candidate genes with domains for DNA methylation in the P. soloecismus genome. Whole-genome bisulfite sequencing revealed DNA methylation in all three cytosine contexts (CpG, CHH, and CHG). While global DNA methylation is low overall (~1.15%), it occurs in appreciable quantities (12.1%) in CpG dinucleotides in a bimodal distribution in all genomic contexts, though terminators contain the greatest number of CpG sites per kilobase. The P. soloecismus genome becomes hypomethylated during the growth cycle in response to nitrogen starvation. Algae cultures were treated daily across the growth cycle with 20 μM 5-aza-2'-deoxycytidine (5AZA) to inhibit propagation of DNA methylation in daughter cells. 5AZA treatment significantly increased optical density and forward and side scatter of cells across the growth cycle (16 days). This increase in cell size and complexity correlated with a significant increase (~66%) in lipid accumulation. Site specific CpG DNA methylation was significantly altered with 5AZA treatment over the time course, though nitrogen starvation itself induced significant hypomethylation in CpG contexts. Genes involved in several biological processes, including fatty acid synthesis, had altered methylation ratios in response to 5AZA; we hypothesize that these changes are potentially responsible for the phenotype of early induction of carbon storage as lipids. This is the first report to utilize epigenetic manipulation strategies to alter algal physiology and phenotype. Collectively, these data suggest these strategies can be utilized to fine-tune metabolic responses, alter growth, and enhance environmental adaption of microalgae for desired outcomes.

5-aza-2'-deoxycytidine↗
DOE OSTI2024

Enhancing Distribution System Resilience: A First-Order Meta-RL Algorithm for Critical Load Restoration

The increasing frequency of extreme events and the integration of distributed energy resources (DERs) into modern grids have elevated the need for resilient and efficient critical load restoration strategies in distribution systems. However, the stochastic nature of renewable DERs, limited energy resource availability and the intricate nonlinearities inherent in complex grid control problem make the problem challenging. Although reinforcement learning (RL) and warm-start RL methods have shown promising results, their performance often falls short in rapidly adapting to new, unseen situations and typically requires exhaustive problem-specific tuning. To address these gaps, we propose a First-Order Meta-based RL (FOM-RL) algorithm within an online framework for adaptive and robust critical load restoration. By harnessing local DERs as the enabling technology, FOM-RL allows the RL agent to swiftly adapt to new unseen scenarios by leveraging previously acquired knowledge of different tasks. Experimental results provide evidence that proposed algorithm learns more efficiently and showcases generalization capabilities across diverse set of operational scenarios. Moreover, a rigorous theoretical analysis yields a tight sublinear regret bound, sensitive to temporal variability, with a task-averaged optimality gap bounded by O(VM+D*/(Tsquare root(M))). These results suggest that optimality improves with task similarity and an increased number of tasks M, reaffirming the efficacy and scalability of the proposed approach in addressing the complexities of critical load restoration in distribution systems.

complexity theory↗
DOE OSTI2022

Improving the oxygen redox reversibility of Li-rich battery cathode materials via Coulombic repulsive interactions strategy

The oxygen redox reaction in lithium-rich layered oxide battery cathode materials generates extra capacity at high cell voltages (i.e., >4.5 V). However, the irreversible oxygen release causes transition metal (TM) dissolution, migration and cell voltage decay. To circumvent these issues, we introduce a strategy for tuning the Coulombic interactions in a model Li-rich positive electrode active material, i.e., Li 1.2 Mn 0.6 Ni 0.2 O 2 . In particular, we tune the Coulombic repulsive interactions to obtain an adaptable crystal structure that enables the reversible distortion of TMO 6 octahedron and mitigates TM dissolution and migration. Moreover, this strategy hinders the irreversible release of oxygen and other parasitic reactions (e.g., electrolyte decomposition) commonly occurring at high voltages. When tested in non-aqueous coin cell configuration, the modified Li-rich cathode material, combined with a Li metal anode, enables a stable cell discharge capacity of about 240 mAh g -1 for 120 cycles at 50 mA g -1 and a slower voltage decay compared to the unmodified Li 1.2 Mn 0.6 Ni 0.2 O 2 .

25 ENERGY STORAGE↗
DOE OSTI2024

Pretraining Billion-Scale Geospatial Foundational Models on Frontier

As AI workloads increase in scope, generalization capability becomes challenging for small task-specific models and their demand for large amounts of labeled training samples increases. On the contrary, Foundation Models (FMs) are trained with internet-scale unlabeled data via self-supervised learning and have been shown to adapt to various tasks with minimal fine-tuning. Although large FMs have demonstrated significant impact in natural language processing and computer vision, efforts toward FMs for geospatial applications have been restricted to smaller size models, as pretraining larger models requires very large computing resources equipped with state-of-the-art hardware accelerators. Current satellite constellations collect 100+TBs of data a day, resulting in images that are billions of pixels and multimodal in nature. Such geospatial data poses unique challenges opening up new opportunities to develop FMs. We investigate billion scale FMs and HPC training profiles for geospatial applications by pretraining on publicly available data. We studied from end-to-end the performance and impact in the solution by scaling the model size. Our larger 3B parameter size model achieves up to 30% improvement in top1 scene classification accuracy when comparing a 100M parameter model. Moreover, we detail performance experiments on the Frontier supercomputer, America's first exascale system, where we study different model and data parallel approaches using PyTorch's Fully Sharded Data Parallel library. Specifically, we study variants of the Vision Transformer architecture (ViT), conducting performance analysis for ViT models with size up to 15B parameters. By discussing throughput and performance bottlenecks under different parallelism configurations, we offer insights on how to leverage such leadership-class HPC resources when developing large models for geospatial imagery applications.

Tsaris, Aristeidis (aris)↗
DOE OSTI2021

Universal dwell time optimization for deterministic optics fabrication

Computer-Controlled Optical Surfacing (CCOS) has been greatly developed and widely used for precision optical fabrication in the past three decades. It relies on robust dwell time solutions to determine how long the polishing tools must dwell at certain points over the surfaces to achieve the expected forms. However, as dwell time calculations are modeled as ill-posed deconvolution, it is always non-trivial to reach a reliable solution that 1) is non-negative, since CCOS systems are not capable of adding materials, 2) minimizes the residual in the clear aperture 3) minimizes the total dwell time to guarantee the stability and efficiency of CCOS processes, 4) can be flexibly adapted to different tool paths, 5) the parameter tuning of the algorithm is simple, and 6) the computational cost is reasonable. In this study, we propose a novel Universal Dwell time Optimization (UDO) model that universally satisfies these criteria. First, the matrix-based discretization of the convolutional polishing model is employed so that dwell time can be flexibly calculated for arbitrary dwell points. Second, UDO simplifies the inverse deconvolution as a forward scalar optimization for the first time, which drastically increases the solution stability and the computational efficiency. Finally, the dwell time solution is improved by a robust iterative refinement and a total dwell time reduction scheme. The superiority and general applicability of the proposed algorithm are verified on the simulations of different CCOS processes. A real application of UDO in improving a synchrotron X-ray mirror using Ion Beam Figuring (IBF) is then demonstrated. The simulation indicates that the estimated residual in the 92.3 mm × 15.7 mm CA can be reduced from 6.32 nm Root Mean Square (RMS) to 0.20 nm RMS in 3.37 min. After one IBF process, the measured residual in the CA converges to 0.19 nm RMS, which coincides with the simulation.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Thiol post-translational modifications modulate allosteric regulation of the OpcA–G6PDH complex through conformational gate control

In cyanobacteria, the redox-sensitive protein OpcA acts as a metabolic switch for G6PDH, enabling rapid adjustment of reducing power generation from glycogen catabolism and thereby precisely regulating carbon flux between anabolic and catabolic pathways. Although redox-sensitive cysteines in OpcA are known to regulate G6PDH, the mechanisms by which redox post-translational modifications (PTMs) on OpcA control G6PDH structure and activity remain unclear. Here, we combine computational modeling with experimental redox proteomics in Synechococcus elongatus PCC 7942 to dissect this mechanism. Experimentally, redox proteome analysis revealed differential redox PTM patterns, particularly on cysteines within the G6PDH-binding site of OpcA. These environmentally sensitive PTM changes at the interface suggest that thiol modifications in this region form a key regulatory node. More broadly, redox proteomics identified site-specific cysteine modifications under light/dark transitions and circadian cycling, linking distinct redox regimes to discrete PTM states. We employed PTM-Psi simulations to show that thiol PTMs near the OpcA–G6PDH interface are critical for allosteric regulation of G6PDH. The thiol PTMs on OpcA affect a putative gate region in G6PDH for substrate ingress and product egress as well as key hydrogen-bond networks within the active site. We infer that PTMs on OpcA tune the conformational landscapes of individual G6PDH subunits toward functionally relevant configurations according to environmental gradients, biasing the enzyme toward catalytically favorable states. Together, our results reveal a molecular mechanism in which thiol PTMs on OpcA modulate G6PDH structure and function through PTM-induced reorganization of conformational dynamics and allosteric communication. These findings demonstrate that PTM-level regulation provides a critical control layer from genotypes to phenotypes that enables cyanobacteria to rapidly adapt to environmental fluctuations through precise metabolic fine-tuning.

Allosteric regulation↗
DOE OSTI2021

FPV Video Adaptation for UAV Collision Avoidance

First person view (FPV) technology for unmanned aerial vehicles (UAVs) provides an immersive experience for pilots and enables various personal and commercial applications such as aerial photography, drone racing, search and rescue operations, agricultural surveillance, and structural inspection. While real time video streaming from a UAV and vision-based collision avoidance strategies have been studied in literature as separate topics, in this paper we tackle collision avoidance in FPV scenarios, taking into account network delays and real time video parameters. We present a theoretical model for obstacle collisions that considers the current communication channel conditions, the real time video parameters, and the UAV's position relative to the closest obstacle. A video adaptation algorithm is then designed, using this metric, to tune the FPV video resolution, number of re-transmission attempts, and the modulation scheme to maximize the probability of avoiding collisions. This algorithm also takes into account specific latency constraints of the application. This video algorithm was evaluated in various scenarios and its ability to respond to both distances to the obstacle as well as the communication channel conditions was demonstrated. It was found that, for the considered scenarios, the performance of the proposed adaptive algorithm was, on an average, 58.63% higher than the closest non-adaptive one in terms of maximizing the probability of avoiding collision. Such collision avoidance strategies could be used to make UAV FPV applications safer and more reliable.

47 OTHER INSTRUMENTATION↗
NASA NTRS2017

An Analysis of the Optimal Control Modification Method Applied to Flutter Suppression

Unlike basic Model Reference Adaptive Control (MRAC)l, Optimal Control Modification (OCM) has been shown to be a promising MRAC modification with robustness and analytical properties not present in other adaptive control methods. This paper presents an analysis of the OCM method, and how the asymptotic property of OCM is useful for analyzing and tuning the controller. We begin with a Lyapunov stability proof of an OCM controller having two adaptive gain terms, then the less conservative and easily analyzed OCM asymptotic property is presented. Two numerical examples are used to show how this property can accurately predict steady state stability and quantitative robustness in the presence of time delay, and relative to linear plant perturbations, and nominal Loop Transfer Recovery (LTR) tuning. The asymptotic property of the OCM controller is then used as an aid in tuning the controller applied to a large scale aeroservoelastic longitudinal aircraft model for flutter suppression. Control with OCM adaptive augmentation is shown to improve performance over that of the nominal non-adaptive controller when significant disparities exist between the controller/observer model and the true plant model.

aeroservoelasti↗
DOE OSTI2023

Accurate prediction of global-density-dependent range-separation parameters based on machine learning

In this work, we develop an accurate and efficient XGBoost machine learning model for predicting the global-density-dependent range-separation parameter, ωGDD, for long-range corrected functional (LRC)-ωPBE. This ωGDDML model has been built using a wide range of systems (11 466 complexes, ten different elements, and up to 139 heavy atoms) with fingerprints for the local atomic environment and histograms of distances for the long-range atomic correlation for mapping the quantum mechanical range-separation values. The promising performance on the testing set with 7046 complexes shows a mean absolute error of 0.001 117 a0−1 and only five systems (0.07%) with an absolute error larger than 0.01 a0−1, which indicates the good transferability of our ωGDDML model. In addition, the only required input to obtain ωGDDML is the Cartesian coordinates without electronic structure calculations, thereby enabling rapid predictions. LRC-ωPBE(ωGDDML) is used to predict polarizabilities for a series of oligomers, where polarizabilities are sensitive to the asymptotic density decay and are crucial in a variety of applications, including the calculations of dispersion corrections and refractive index, and surpasses the performance of all other popular density functionals except for the non-tuned LRC-ωPBE. Finally, LRC-ωPBE (ωGDDML) combined with (extended) symmetry-adapted perturbation theory is used in calculating noncovalent interactions to further show that the traditional ab initio system-specific tuning procedure can be bypassed. The present study not only provides an accurate and efficient way to determine the range-separation parameter for LRC-ωPBE but also shows the synergistic benefits of fusing the power of physically inspired density functional LRC-ωPBE and the data-driven ωGDDML model.

Chemistry↗
DOE OSTI2024

Learning Robust Marking Policies for Adaptive Mesh Refinement

Here in this work, we revisit the marking decisions made in the standard adaptive finite element method (AFEM). Experience shows that a naïve marking policy leads to inefficient use of computational resources for adaptive mesh refinement (AMR). Consequently, using AMR in practice often involves ad-hoc or time-consuming offline parameter tuning to set appropriate parameters for the marking subroutine. To address these practical concerns, we recast AMR as a Markov decision process in which refinement parameters can be selected on-the-fly at run time, without the need for pre-tuning by expert users. In this new paradigm, the refinement parameters are also chosen adaptively via a marking policy that can be optimized using methods from reinforcement learning. We use the Poisson equation to demonstrate our techniques on h- and hp-refinement benchmark problems, and our experiments suggest that superior marking policies remain undiscovered for many classical AFEM applications. Furthermore, an unexpected observation from this work is that marking policies trained on one family of PDEs are sometimes robust enough to perform well on problems far outside the training family. For illustration, we show that a simple hp-refinement policy trained on 2D domains with only a single re-entrant corner can be deployed on far more complicated 2D domains, and even 3D domains, without significant performance loss. For reproduction and broader adoption, we accompany this work with an open-source implementation of our methods.

97 MATHEMATICS AND COMPUTING↗