Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Learning Framework”

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 1,135 records · Page 63
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Physics-informed latent neural operator for real-time predictions of time-dependent parametric PDEs

Deep operator network (DeepONet) has shown significant promise as surrogate models for systems governed by partial differential equations (PDEs), enabling accurate mappings between infinite-dimensional function spaces. However, when applied to systems with high-dimensional input-output mappings arising from large numbers of spatial and temporal collocation points, these models often require heavily overparameterized networks, leading to long training times. Latent DeepONet addresses some of these challenges by introducing a two-step approach: first learning a reduced latent space using a separate model, followed by operator learning within this latent space. While efficient, this method is inherently data-driven and lacks mechanisms for incorporating physical laws, limiting its robustness and generalizability in data-scarce settings. Here, in this work, we propose PI-Latent-NO, a physics-informed latent neural operator framework that integrates governing physics directly into the learning process. Our architecture features two coupled DeepONets trained end-to-end: a Latent-DeepONet that learns a low-dimensional representation of the solution, and a Reconstruction-DeepONet that maps this latent representation back to the physical space. By embedding PDE constraints into the training via automatic differentiation, our method eliminates the need for labeled training data and ensures physics-consistent predictions. The proposed framework is both memory and compute-efficient, exhibiting near-constant scaling with problem size and demonstrating significant speedups over traditional physics-informed operator models. We validate our approach on a range of parametric PDEs, showcasing its accuracy, scalability, and suitability for real-time prediction in complex physical systems.

Latent representations↗
NASA NTRS2006

Implementing Scientific Simulation Codes Highly Tailored for Vector Architectures Using Custom Configurable Computing Machines

The motivation for this work comes from an observation that amidst the push for Massively Parallel (MP) solutions to high-end computing problems such as numerical physical simulations, large amounts of legacy code exist that are highly optimized for vector supercomputers. Because re-hosting legacy code often requires a complete re-write of the original code, which can be a very long and expensive effort, this work examines the potential to exploit reconfigurable computing machines in place of a vector supercomputer to implement an essentially unmodified legacy source code. Custom and reconfigurable computing resources could be used to emulate an original application's target platform to the extent required to achieve high performance. To arrive at an architecture that delivers the desired performance subject to limited resources involves solving a multi-variable optimization problem with constraints. Prior research in the area of reconfigurable computing has demonstrated that designing an optimum hardware implementation of a given application under hardware resource constraints is an NP-complete problem. The premise of the approach is that the general issue of applying reconfigurable computing resources to the implementation of an application, maximizing the performance of the computation subject to physical resource constraints, can be made a tractable problem by assuming a computational paradigm, such as vector processing. This research contributes a formulation of the problem and a methodology to design a reconfigurable vector processing implementation of a given application that satisfies a performance metric. A generic, parametric, architectural framework for vector processing implemented in reconfigurable logic is developed as a target for a scheduling/mapping algorithm that maps an input computation to a given instance of the architecture. This algorithm is integrated with an optimization framework to arrive at a specification of the architecture parameters that attempts to minimize execution time, while staying within resource constraints. The flexibility of using a custom reconfigurable implementation is exploited in a unique manner to leverage the lessons learned in vector supercomputer development. The vector processing framework is tailored to the application, with variable parameters that are fixed in traditional vector processing. Benchmark data that demonstrates the functionality and utility of the approach is presented. The benchmark data includes an identified bottleneck in a real case study example vector code, the NASA Langley Terminal Area Simulation System (TASS) application.

Rutishauser, David↗
DOE OSTI2024

Progressive transfer learning for advancing machine learning-based reduced-order modeling

Abstract To maximize knowledge transfer and improve the data requirement for data-driven machine learning (ML) modeling, a progressive transfer learning for reduced-order modeling (p-ROM) framework is proposed. A key concept of p-ROM is to selectively transfer knowledge from previously trained ML models and effectively develop a new ML model(s) for unseen tasks by optimizing information gates in hidden layers. The p-ROM framework is designed to work with any type of data-driven ROMs. For demonstration purposes, we evaluate the p-ROM with specific Barlow Twins ROMs (p-BT-ROMs) to highlight how progress learning can apply to multiple topological and physical problems with an emphasis on a small training set regime. The proposed p-BT-ROM framework has been tested using multiple examples, including transport, flow, and solid mechanics, to illustrate the importance of progressive knowledge transfer and its impact on model accuracy with reduced training samples. In both similar and different topologies, p-BT-ROM achieves improved model accuracy with much less training data. For instance, p-BT-ROM with four-parent (i.e., pre-trained models) outperforms the no-parent counterpart trained on data nine times larger. The p-ROM framework is poised to significantly enhance the capabilities of ML-based ROM approaches for scientific and engineering applications by mitigating data scarcity through progressively transferring knowledge.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2006

QUICR-learning for Multi-Agent Coordination

Coordinating multiple agents that need to perform a sequence of actions to maximize a system level reward requires solving two distinct credit assignment problems. First, credit must be assigned for an action taken at time step t that results in a reward at time step t > t. Second, credit must be assigned for the contribution of agent i to the overall system performance. The first credit assignment problem is typically addressed with temporal difference methods such as Q-learning. The second credit assignment problem is typically addressed by creating custom reward functions. To address both credit assignment problems simultaneously, we propose the "Q Updates with Immediate Counterfactual Rewards-learning" (QUICR-learning) designed to improve both the convergence properties and performance of Q-learning in large multi-agent problems. QUICR-learning is based on previous work on single-time-step counterfactual rewards described by the collectives framework. Results on a traffic congestion problem shows that QUICR-learning is significantly better than a Q-learner using collectives-based (single-time-step counterfactual) rewards. In addition QUICR-learning provides significant gains over conventional and local Q-learning. Additional results on a multi-agent grid-world problem show that the improvements due to QUICR-learning are not domain specific and can provide up to a ten fold increase in performance over existing methods.

Agogino, Adrian K.↗
DOE OSTI2024

Machine Learning Potentials with the Iterative Boltzmann Inversion: Training to Experiment

Methodologies for training machine learning potentials (MLPs) with quantum-mechanical simulation data have recently seen tremendous progress. Experimental data have a very different character than simulated data, and most MLP training procedures cannot be easily adapted to incorporate both types of data into the training process. Here, we investigate a training procedure based on iterative Boltzmann inversion that produces a pair potential correction to an existing MLP using equilibrium radial distribution function data. By applying these corrections to an MLP for pure aluminum based on density functional theory, we observe that the resulting model largely addresses previous overstructuring in the melt phase. Interestingly, the corrected MLP also exhibits improved performance in predicting experimental diffusion constants, which are not included in the training procedure. The presented method does not require autodifferentiating through a molecular dynamics solver and does not make assumptions about the MLP architecture. Our results suggest a practical framework for incorporating experimental data into machine learning models to improve the accuracy of molecular dynamics simulations.

36 MATERIALS SCIENCE↗
DOE OSTI2022

HyMARC Seedling: Optimized Hydrogen Adsorbents via Machine Learning and Crystal Engineering

This final report is presented in two parts: In Part 1, the H 2 capacities of a diverse set of 918,734 metal-organic frameworks (MOFs) sourced from 19 databases is predicted via machine learning (ML). Using only 7 structural features as input, ML identifies 8,282 MOFs with the potential to exceed the capacities of state-of-the-art materials. The identified MOFs are predominantly hypothetical compounds having low densities (<0.31 g/cm3) in combination with high surface areas (>5,300 m2/g), void fractions ($0.90), and pore volumes (>3.3 cm3/g). The relative importance of the input features are characterized, and dependencies on the ML algorithm and training set size are quantified. The most important features for predicting H 2 uptake are pore volume (for gravimetric capacity) and void fraction (for volumetric capacity). The ML models are available on the web, allowing for rapid and accurate predictions of the hydrogen capacities of MOFs from limited structural data; the simplest models require only a single crystallographic feature. In part 2, ways to improve the poor powder packing density of MOFs is discussed. More specifically, a strategy that improves packing efficiency and volumetric hydrogen gas storage density dramatically through engineered morphologies and controlled-crystal size distributions is presented that holds promise for maximizing storage capacity for a given MOF. The packing density improvement, demonstrated for the benchmark sorbent MOF-5, leads to a significant enhancement of volumetric hydrogen storage performance relative to commercial MOF-5. System model projections demonstrate that engineering of crystal morphology/size or use of a bimodal distribution of cubic crystal sizes in tandem with system optimization can surpass the 25 g/L volumetric capacity of a typical 700 bar compressed storage system and exceed the DOE targets 2020 volumetric capacity (30 g/L). Finally, a critical link between improved powder packing density and reduced damage upon compaction is revealed leading to sorbents with both high surface area and high density.

08 HYDROGEN↗
DOE OSTI2022

A Hybrid Reinforcement Learning-MPC Approach for Distribution System Critical Load Restoration: Preprint

This paper proposes a hybrid control approach for distribution system critical load restoration, combining deep reinforcement learning (RL) and model predictive control (MPC) aiming at maximizing total restored load following an extreme event. RL determines a policy for quantifying operating reserve requirements, thereby hedging against uncertainty, while MPC models grid operations incorporating RL policy actions, i.e., the reserve requirement, renewable (wind and solar) power predictions, and load demand forecasts. We formulate the reserve requirement determination problem as a sequential decision making problem based on the Markov Decision Process (MDP) and design an RL learning environment based on the OpenAI Gym framework and MPC. The RL agent reward and MPC objective function aim to maximize and monotonically increase total restored load and minimize load shedding and renewable power curtailment. The RL algorithm is trained off-line using historical forecast of renewable generation and load demand. The method is tested using a modified IEEE 13-bus distribution test feeder containing wind turbine, photovoltaic, microturbine and battery. Case studies demonstrated that the proposed method outperforms other operating reserve determination methods.

distribution system↗
DOE OSTI2022

A Hybrid Reinforcement Learning-MPC Approach for Distribution System Critical Load Restoration

This paper proposes a hybrid control approach for distribution system critical load restoration, combining deep reinforcement learning (RL) and model predictive control (MPC) aiming at maximizing total restored load following an extreme event. RL determines a policy for quantifying operating reserve requirements, thereby hedging against uncertainty, while MPC models grid operations incorporating RL policy actions (i.e., reserve requirements), renewable (wind and solar) power predictions, and load demand forecasts. We formulate the reserve requirement determination problem as a sequential decision-making problem based on the Markov Decision Process (MDP) and design an RL learning environment based on the OpenAI Gym framework and MPC simulation. The RL agent reward and MPC objective function aim to maximize and monotonically increase total restored load and minimize load shedding and renewable power curtailment. The RL algorithm is trained offline using a historical forecast of renewable generation and load demand. The method is tested using a modified IEEE 13-bus distribution test feeder containing wind turbine, photovoltaic, microturbine, and battery. Case studies demonstrated that the proposed method outperforms other policies with static operating reserves.

distribution system↗
DOE OSTI2020

Data Requirements for Application of Risk-Based Dynamic Contingency Analysis to Evaluate Hurricane Impact to Electrical Infrastructure in Puerto Rico

This paper presents a risk-based dynamic contingency analysis framework that was used to evaluate the hurricane impact to electrical infrastructure in Puerto Rico. PNNL developed a scalable risk-based framework for identifying high-voltage transmission resilience improvements by classifying and prioritizing high-risk power grid contingencies (system failures) under hurricane impact. The risk-based framework is founded on grid outage definitions with their associated probabilities of occurrence from hurricane events, in combination with an impact assessment derived from detailed dynamic cascading failure analysis. This paper focuses on a discussion around data requirements for transmission resilience planning for hurricane events, derived from the development of the risk-based framework and its application to Puerto Rico. This paper launches an important first step in encouraging the engineering community and power system industry to move towards establishing resilience planning as a routine practice. Since actual results for Puerto Rico contain sensitive information, sample simulation results will be used to illustrate the data requirements and risk-based dynamic cascading framework on the Puerto Rico power grid, as well as demonstrate the potential for such a simulation framework. The paper includes a discussion on the lessons learned, importance and need for improved datasets that are not usually considered in traditional power system planning. The paper will also elaborate on how the scalable simulation framework and datasets might be expanded to larger footprints and leveraged for modelling other types of natural disasters.

DCAT, Puerto Rico, hurricane, Power System Stabili↗
DOE OSTI2025

Physics-guided dual implicit neural representations for source separation

Significant challenges exist in efficient data analysis of most advanced experimental and observational techniques because the collected signals often include unwanted contributions, such as background and signal distortions, that can obscure the physically relevant information of interest. To address this, we have developed a self-supervised machine-learning approach for source separation using a dual implicit neural representation framework that jointly trains two neural networks: one for approximating distortions of the physical signal of interest and the other for learning the effective background contribution. Our method learns directly from the raw data by minimizing a reconstruction-based loss function without requiring labeled data or pre-defined dictionaries. We demonstrate the effectiveness of our framework by considering a challenging case study involving large-scale simulated, as well as experimental, momentum-energy-dependent inelastic neutron scattering data in a four-dimensional parameter space, characterized by heterogeneous background contributions and unknown distortions to the target signal. The method is found to successfully separate physically meaningful signals from a complex or structured background even when the signal characteristics vary across all four dimensions of the parameter space. An analytical approach that informs the choice of the regularization parameter is presented. Our method offers a versatile framework for addressing source separation problems across diverse domains, ranging from superimposed signals in astronomical measurements to structural features in biomedical image reconstructions.

47 OTHER INSTRUMENTATION↗
DOE OSTI2025

Optimizing on-ramp merging for connected and automated vehicles: A hierarchical approach using deep reinforcement learning and optimal control

On-ramp merging for Connected and Automated Vehicles (CAVs) presents significant challenges in dynamic traffic environments. Traditional methods and recent learning-based approaches often fail to simultaneously address decision-making complexity and execution precision under fluctuating conditions. This study introduces a novel hierarchical framework that combines: (1) a high-level Deep Reinforcement Learning (DRL) module that coordinates merging sequences through Virtual Traffic Signals (VTS) with Yield/Green phases and (2) a low-level optimal controller generating collision-free speed trajectories via pseudospectral convex optimization. A convolutional autoencoder compresses high-dimensional traffic states to enhance responsiveness. Extensive simulations demonstrate a 12.5% improvement in mainline throughput a 28% reduction in emergency braking events, and 31.66% lower fuel consumption compared to baseline methods. Furthermore, the framework’s effectiveness in coordinating CAV merges highlights its potential for real-world deployment. Future work will extend validation to multi-lane scenarios with mixed traffic and large-scale multiple merging points.

Connected and automated vehicles↗
NASA NTRS2001

Implementation of a General Real-Time Visual Anomaly Detection System Via Soft Computing

The intelligent visual system detects anomalies or defects in real time under normal lighting operating conditions. The application is basically a learning machine that integrates fuzzy logic (FL), artificial neural network (ANN), and generic algorithm (GA) schemes to process the image, run the learning process, and finally detect the anomalies or defects. The system acquires the image, performs segmentation to separate the object being tested from the background, preprocesses the image using fuzzy reasoning, performs the final segmentation using fuzzy reasoning techniques to retrieve regions with potential anomalies or defects, and finally retrieves them using a learning model built via ANN and GA techniques. FL provides a powerful framework for knowledge representation and overcomes uncertainty and vagueness typically found in image analysis. ANN provides learning capabilities, and GA leads to robust learning results. An application prototype currently runs on a regular PC under Windows NT, and preliminary work has been performed to build an embedded version with multiple image processors. The application prototype is being tested at the Kennedy Space Center (KSC), Florida, to visually detect anomalies along slide basket cables utilized by the astronauts to evacuate the NASA Shuttle launch pad in an emergency. The potential applications of this anomaly detection system in an open environment are quite wide. Another current, potentially viable application at NASA is in detecting anomalies of the NASA Space Shuttle Orbiter's radiator panels.

Dominguez, Jesus A.↗
DOE OSTI2025

Even Higher-Level Synthesis: An Exploration of AI Hardware Accelerators using HLS4ML

With the rise of artificial intelligence, the popularization of deep learning, and a constantly evolving industry, the demand for flexible and efficient tools has never been greater. As algorithms grow more complex, their runtime and energy consumption increase exponentially. Customized hardware accelerators, long used for specific mathematical operations, remain essential for managing modern applications' computational and power demands. Hardware accelerators can speed up complex computations by orders of magnitude, but their manual design and verification processes are often challenging and time-consuming. High-Level Synthesis (HLS) provides a solution by transforming high-level algorithm descriptions, typically written in C++ or SystemC, into synthesizable RTL suitable for hardware implementation. This approach reduces development time for RTL engineers while offering flexibility beyond what traditional handwritten RTL can provide. We extended this capability to the machine-learning domain with the open-source framework hls4ml, which allows neural networks trained in Python frameworks like Tensorflow or PyTorch to be synthesized into efficient hardware representations for the traditional FPGA and ASIC flows. This breakthrough addresses the growing need for reduced design turnaround and easy verification of ML hardware accelerators with low latency and power efficiency constraints. During this tutorial, we will demonstrate how Python complements HLS by simplifying the ML design process, bridging the gap between software and hardware development. Attendees will explore how we translate neural networks modeled in Python into fixed-point C++ models suitable for HLS workflows. We will dive into strategies like Value-Range Analysis and Quantization-Aware Training, which optimize these designs for deployment and evaluate their accuracy, power consumption, and energy efficiency. To exemplify these concepts, experts from Fermilab will share their experiences applying this technology to high-energy physics experiments, where real-time, low-latency processing is critical. Over the years, Fermilab engineers have demonstrated how deep neural networks, optimized for hardware using hls4ml, can meet the stringent requirements of trigger systems at the CERN Large Hadron Collider. These systems rely on rapid decision-making to process immense data volumes while retaining only the most relevant events for further analysis. The application of hls4ml has also been extended to innovative technologies like smart pixel arrays. These smart pixels integrate ML inference capabilities directly into sensor devices, enabling localized data processing at the pixel level. This approach drastically reduces the need to transmit raw data to external processing units, significantly decreasing power consumption and latency. By embedding neural networks within the pixel architecture, the smart pixels can identify and prioritize relevant data in real time, providing a highly efficient solution for edge computing in scenarios such as particle detectors and imaging systems. Fermilab's work highlights the potential of hardware-accelerated ML in scenarios where both speed and power efficiency are mission-critical. Through this tutorial, attendees will gain valuable insights into the challenges and solutions of deploying ML in hardware. Understanding how HLS and hls4ml streamline the development of neural network-based hardware accelerators is fundamental for the industry's future. Participants will learn how these technologies are shaping the future of AI and scientific computing.

Di Guglielmo, Giuseppe [Fermilab]↗
DOE OSTI2025

Efficient and assured reinforcement learning-based building HVAC control with heterogeneous expert-guided training

Abstract Building heating, ventilation, and air conditioning (HVAC) systems account for nearly half of building energy consumption and $$20\%$$ of total energy consumption in the US. Their operation is also crucial for ensuring the physical and mental health of building occupants. Compared with traditional model-based HVAC control methods, the recent model-free deep reinforcement learning (DRL) based methods have shown good performance while do not require the development of detailed and costly physical models. However, these model-free DRL approaches often suffer from long training time to reach a good performance, which is a major obstacle for their practical deployment. In this work, we present a systematic approach to accelerate online reinforcement learning for HVAC control by taking full advantage of the knowledge from domain experts in various forms . Specifically, the algorithm stages include learning expert functions from existing abstract physical models and from historical data via offline reinforcement learning, integrating the expert functions with rule-based guidelines, conducting training guided by the integrated expert function and performing policy initialization from distilled expert function. Moreover, to ensure that the learned DRL-based HVAC controller can effectively keep room temperature within the comfortable range for occupants, we design a runtime shielding framework to reduce the temperature violation rate and incorporate the learned controller into it. Experimental results demonstrate up to 8.8 X speedup in DRL training from our approach over previous methods, with low temperature violation rate.

Xu, Shichao↗
DOE OSTI2024

Building an ab initio solvated DNA model using Euclidean neural networks

Accurately modeling large biomolecules such as DNA from first principles is fundamentally challenging due to the steep computational scaling of ab initio quantum chemistry methods. This limitation becomes even more prominent when modeling biomolecules in solution due to the need to include large numbers of solvent molecules. We present a machine-learned electron density model based on a Euclidean neural network framework that includes a built-in understanding of equivariance to model explicitly solvated double-stranded DNA. By training the machine learning model using molecular fragments that sample the key DNA and solvent interactions, we show that the model predicts electron densities of arbitrary systems of solvated DNA accurately, resolves polarization effects that are neglected by classical force fields, and captures the physics of the DNA-solvent interaction at the ab initio level.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Chapter 7: Learning Stable Local Volt/Var Controllers in Distribution Grids

This chapter describes a framework to synthesize provably stable local Volt/Var controllers for distributed energy resources (DERs) in power distribution grids (DGs). The goal is to control the reactive power injections of DERs to improve the system performance as quantified by a generic optimal reactive power flow (ORPF) problem. To achieve this, we jointly design for each DER the control function, which prescribes the reactive power update rule, and the equilibrium function, which approximates the ORPF solutions from local measurements of voltages and powers. We provide conditions on the equilibrium functions and the control parameters ensuring the stability of the closed-loop system. In particular, we discuss the trade-offs between each set of conditions accounting for practical considerations, like fully exploiting the DERs' generation capabilities and reducing the optimality gap. These conditions are then translated into learning constraints on the neural networks' parameters that are enforced in the training phase. We validate our framework with numerical simulations on the IEEE 37-bus network and through a comparison with an optimized version of standard piece wise linear control rules.

closed-loop asymptotic stability↗