SEARCH · Search NASA
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.
This presentation is during a course on Machine Learning in Mechanics at the Society of Experimental Mechanics Annual Conference 2025.
Engineering analyses on Surveyor lunar dust particle detector instrumentation, and ground support equipment
HPC datacenters offer a backbone to the modern digital society. Increasingly, they run Machine Learning (ML) jobs next to generic, compute-intensive workloads, supporting science, business, and other decision-making processes. However, understanding how ML jobs impact the operation of HPC datacenters, relative to generic jobs, remains desirable but understudied. In this work, we leverage long-term operational data, collected from a national-scale production HPC datacenter, and statistically compare how ML and generic jobs can impact the performance, failures, resource utilization, and energy consumption of HPC datacenters. Our study provides key insights, e.g., ML-related power usage causes GPU nodes to run into temperature limitations, median/mean runtime and failure rates are higher for ML jobs than for generic jobs, both ML and generic jobs exhibit highly variable arrival processes and resource demands, significant amounts of energy are spent on unsuccessfully terminating jobs, and concurrent jobs tend to terminate in the same state. We open-source our cleaned-up data traces on Zenodo (https://doi. org/10.5281/zenodo.13685426), and provide our analysis toolkit as software hosted on GitHub (https://github.com/atlarge-research/2024-icpads-hpc-workload-characterization). This study offers multiple benefits for data center administrators, who can improve operational efficiency, and for researchers, who can further improve system designs, scheduling techniques, etc.
High Performance Computing (HPC) applications rely heavily on code optimizations to achieve good performance on modern CPU and GPU architectures. Traditional Machine Learning auto-tuning approaches have demonstrated success in exploring high-dimensional spaces, but they often require expensive compile-run evaluations and lack adaptability for large HPC applications. The recent advances in Large Language Models (LLMs) and Agentic AI systems raise intriguing questions about the potential of these approaches to address specific optimization methodologies. This work aims to answer an essential question for the HPC community: “How Agentic AI Systems Compare to Traditional ML Autotuning Techniques?” To address this question, we present a comparative analysis between a traditional ML-based optimization approach and an Agentic AI system, evaluating their respective capabilities and limitations for loop-level optimization. In addition, we introduced a new Agentic AI system named LoopGen-AI using three different Large Language Models: GPT-4.1, Claude 4.0, and Gemini 2.5. A key finding is that LoopGen-AI achieves competitive per-formance with only a few program runs, the reasoning logs from the agents revealed that their decisions rely heavily on the combination of semantic understanding of the target kernel with dynamic feedback from the environment, highlighting a promising new dimension in performance tuning. In contrast, ML-based autotuners focus on statistical exploration, and require orders of magnitude more runs to reach peak performance. Additionally, our analysis shows that prompt engineering, particularly using Persona + Context Manager patterns, significantly impacts the effectiveness of Agentic AI. Our results indicate that while Agentic AI systems are not yet a complete replacement for ML-based autotuners, it can effectively complement traditional methods.
In this paper, we propose a method to perform empirical analysis of the loss landscape of machine learning (ML) models. The method is applied to two ML models for scientific sensing, which necessitates quantization to be deployed and are subject to noise and perturbations due to experimental conditions. Our method allows assessing the robustness of ML models to such effects as a function of quantization precision and under different regularization techniques -- two crucial concerns that remained underexplored so far. By investigating the interplay between performance, efficiency, and robustness by means of loss landscape analysis, we both established a strong correlation between gently-shaped landscapes and robustness to input and weight perturbations and observed other intriguing and non-obvious phenomena. Our method allows a systematic exploration of such trade-offs a priori, i.e., without training and testing multiple models, leading to more efficient development workflows. This work also highlights the importance of incorporating robustness into the Pareto optimization of ML models, enabling more reliable and adaptive scientific sensing systems.
Accurately predicting solid electrolyte interphase (SEI) formation requires explicitly resolving the electric double layer (EDL) structure, which deviates significantly from that of the bulk electrolyte. Although an established molecular dynamics (MD) and Density Functional Theory (DFT) framework can model SEI formation by evaluating reduction reactions of local clusters in the EDL, it suffers from a combinatorial computational bottleneck. To overcome this limitation, we introduce a machine-learning-accelerated simulation workflow (MD–DFT–ML), integrating a gradient-boosted regression model trained on EDL composition data to efficiently predict reduction potentials. We apply this framework to seven fluorinated electrolytes comprising fluorinated anions, a fluorinated ester solvent, two types of diluent (ion-solvating ester vs. non-solvating ether), and an FEC additive. The analysis shows that the EDL selectively accumulates cation-binding species; consequently, the non–cation-binding ether diluent rarely enters the EDL and makes minimal contributions to SEI formation. DFT calculations on statistically representative EDL clusters provide reduction potentials and fluorine-release pathways, while the ML model, which substantially reduces the DFT workload, predicts cluster reduction energies with a mean absolute error of 0.1 eV. The combined MD–DFT–ML approach also quantifies contributions from different sources to LiF formation in the SEI. This methodology establishes a generalizable route for multiscale modeling electrolyte and interphase design for next-generation electrochemical energy-storage systems.
LIBS−ML framework for real time feedstock characterization during continuous conveyor transport Heterogeneous waste derived feedstocks (e.g., waste coal, biomass and blends) introduce rapid variability in heating value and ash chemistry that affect gasifier operation, yet conventional laboratory characterization techniques are too slow to support proactive control. To address this gap, this study reports on an online, in situ, dynamic characterization framework that couple’s laser-induced breakdown spectroscopy (LIBS) with leakage safe machine learning (ML) regression to deliver real time, decision quality predictions of gasifier relevant properties. A controlled sample matrix spanning two different waste coals, two different biomasses, and engineered blends under two particle size conditions were constructed and benchmarked using standardized laboratory analyses for proximate/ultimate properties and ash composition. LIBS spectra were acquired dynamically as material flowed on a conveyor belt, using high energy 1064 nm laser ablation and shot averaging to improve repeatability and precision. Supervised regression models (multi layer perceptron (MLP) /artificial neural network (ANN), random forest (RF), and support vector regression (SVR)) and an optimized weighted ensemble were trained on emission line feature sets using nested cross validation with Bayesian hyperparameter tuning and validated against an independent hold out set. The proposed LIBS−ML workflow achieves near laboratory predictive fidelity across parametric targets (including higher heating value (HHV), ash content, fixed carbon, sulfur, major ash forming oxides, and initial deformation temperature (IDT)), with the weighted ensemble providing a robust default predictor under dynamic measurement conditions. These results demonstrate a practical pathway for real time feedstock characterization that can enable feedforward adjustments and more resilient gasifier operation for variable quality waste derived fuels.
Abstract Machine‐learning (ML) parameterizations of subgrid processes (here of turbulence, convection, and radiation) may one day replace conventional parameterizations by emulating high‐resolution physics without the cost of explicit simulation. However, uncertainty about the relationship between offline and online performance (i.e., when integrated with a large‐scale general circulation model) hinders their development. Much of this uncertainty stems from limited sampling of the noisy, emergent effects of upstream ML design decisions on downstream online hybrid simulation. Our work rectifies the sampling issue via the construction of a semi‐automated, end‐to‐end pipeline for size ensembles of hybrid simulations, revealing important nuances in how systematic reductions in offline error manifest in changes to online error and online stability. For example, removing dropout and switching from a Mean Squared Error to a Mean Absolute Error loss both reduce offline error, but they have opposite effects on online error and online stability. Other design decisions, like incorporating memory, converting moisture input from specific humidity to relative humidity, using batch normalization, and training on multiple climates do not come with any such compromises. Finally, we show that ensemble sizes of may be necessary to reliably detect causally relevant differences online. By enabling rapid online experimentation at scale, we can empirically settle debates regarding subgrid ML parameterization design that would have otherwise remained unresolved in the noise.
Subgrid machine-learning (machine learning [ML]) parameterizations have the potential to introduce a new generation of climate models that incorporate the effects of higher-resolution physics without incurring the prohibitive computational cost associated with more explicit physics-based simulations. However, important issues, ranging from online instability to inconsistent online performance, have limited their operational use for long-term climate projections. To more rapidly drive progress in solving these issues, domain scientists and ML researchers opened up the offline aspect of this problem to the broader ML and data science community with the release of ClimSim, a NeurIPS Data sets and Benchmarks publication, and an associated Kaggle competition. This paper reports on the downstream results of the Kaggle competition by coupling emulators inspired by the winning teams' architectures to an interactive climate model (including full cloud microphysics, a regime historically prone to online instability) and systematically evaluating their online performance. Our results demonstrate that online stability in the low-resolution real-geography setting is reproducible across multiple diverse architectures, which we consider a key milestone. All tested architectures exhibit strikingly similar offline and online biases, though their responses to architecture-agnostic design choices (e.g., expanding the list of input variables) can differ significantly. Multiple Kaggle-inspired architectures achieve state-of-the-art results on certain metrics such as zonal mean bias patterns and global Root Mean Squared Error, indicating that crowdsourcing the essence of the offline problem is one path to improving online performance in hybrid physics-AI climate simulation.
DAFTAR-ML is a specialized machine-learning pipeline that identifies relevant features based on their relationship to a target variable. Many ML pipelines focus solely on prediction, and feature ranking is often absent or lacks robust statistical methods. DAFTAR-ML performs its tasks with this outcome in mind. Model training is robust, using nested cross-validation and hyperparameter tuning. Instead of relying on native feature-importance scores, it employs SHAP (SHapley Additive exPlanations) to quantify feature importance. The pipeline also produces comprehensive results, including publication-quality visualizations.
Real-time and adaptive plasma control is crucial for robust tokamak operation, requiring sensitivity and tolerance measurements of the plasma state. This paper presents the implementation of an integrated real-time plasma monitoring framework on the DIII-D tokamak to support advanced control approaches, including machine-learning (ML) methods. The system is built on the SHIELD framework, a high-performance modular architecture that provides a unified pipeline for integrating diverse diagnostics. The framework leverages high-bandwidth digitizers, fast numerical processing, and deterministic, low-latency interconnects to stream high-fidelity data from diagnostics such as electron cyclotron emission (ECE), beam emission spectroscopy (BES), CO interferometers, and a visible tangential divertor camera (TangTV). The system’s validity is demonstrated through direct comparisons of real-time and offline data. Furthermore, we present two key applications of the developed plasma monitoring system with ML-based plasma control strategies, including real-time divertor detachment and active Alfvén Eigenmode control. As a result, this work presents a robust and scalable approach for integrating high-frequency, multidimensional diagnostics into advanced control algorithms for future fusion devices.
The heterogeneity in the composition of municipal solid wastes (MSW) poses significant challenges in the production of biofuel and bioproducts. This research aims to enhance the accuracy and efficiency of waste analysis and characterization by introducing a fast characterization approach for MSW-derived refuse-derived fuels (RDF) by combining Laser-Induced Breakdown Spectroscopy (LIBS) with advanced machine learning (ML) techniques. The approach combines data pre-processing of LIBS spectra of RDF, and the development of ML models trained on domain and theory-based spectral features for predicting process parameters. These models are adept at predicting key process parameters like High Heating Value (HHV), carbon content, and volatile matter. This approach can achieve an average RRMSE of 2.13% and R 2 of 0.98 or higher for all considered parameters on testing data. This work demonstrates significant potential for improving waste sorting, processing efficiency, and environmental compliance over traditional labor- and time-intensive laboratory waste analysis and characterization.
Accurate and low-latency quantum state discrimination is essential for protocols involving mid-circuit measurement (MCM) and conditional feed-forward. In superconducting quantum systems, conventional readout pipelines transfer measurement data to host processors for post-processing, introducing millisecond-scale delays that far exceed qubit coherence times. To overcome this bottleneck, we present an in-situ machine learning (ML) inference engine implemented on an FPGA for real-time quantum state discrimination. Our design performs inference directly on digitized readout signals with 40 ns latency, supports both qubit and qutrit readout, and enables conditional operations without host-side intervention. This capability is critical for MCM and for feedback-driven protocols such as quantum error correction. We validate the system on superconducting transmon hardware, demonstrating robust discrimination fidelity across multiple qubit and qutrit channels. We further demonstrate conditional qutrit logic driven by FPGA-resident classification, highlighting the potential of low-latency ML-on-FPGA control for NISQ applications and scalable fault-tolerant quantum computing.
Detecting quenches in superconducting (SC) magnets during training is a challenging process that involves capturing physical events that occur at different frequencies and appear as various signal features. These events may be correlated across instrumentation type, thermal cycle, and ramp. These events together build a more complete picture of continuous processes occurring in the magnet, and may allow us to flag potential precursors for quench detection. We present our work on building an automatic machine learning (ML) based quench detection system. We build upon our existing work on unsupervised auto-encoders for acoustic sensors and quench antenna (QA) by first establishing a supervised ML training pipeline. We show the results of an event tagging, analysis, and simulation framework on our QA and acoustic data which are used concurrently to build a training dataset for a supervised implementation. We then show how this supervised training can be used as a prior in a semi-supervised framework and compare this to the unsupervised neural network auto-encoder performance.This allows us to have a more concrete understanding of the performance of our algorithms relative to physical events occurring in the magnet, and also provides a baseline software tool to generically evaluate our quench prediction autoencoders under completely unsupervised, supervised, and semi-supervised training conditions.
Here, we introduce a hybrid computational framework that merges HPC-based numerical solvers with physics-informed ML surrogates for efficient modeling of magnon-photon interactions. By running short-duration, high-fidelity Maxwell-LLG simulations and feeding their results into an ML model, we substantially cut simulation time while achieving accurate predictions across larger spatiotemporal domains.
We present results on extending the strong lens discovery space down to much smaller Einstein radii ( θ E ≲ $0^{_{''}}_{^.}03$) and much lower halo mass (M halo < 10 11 M ⊙ ) through the combination of JWST observations and machine learning (ML) techniques. First, we forecast detectable strong lenses with JWST using CosmoDC2 as the lens catalog, and a source catalog down to 29th magnitude. By further incorporating the VELA hydrodynamical simulations of high-redshift galaxies, we simulate strong lenses. We train a ResNet on these images, achieving near-100% completeness and purity for “conventional” strong lenses ( θ E ≳ $0^{_{''}}_{^.}05$), applicable to JWST, the Hubble Space Telescope (HST), the Roman Space Telescope, and Euclid VIS. For the first time, we also search for very low halo mass strong lenses (M halo < 10 11 M ⊙ ) in simulations, with θ E << $0^{_{''}}_{^.}05$, down to the best resolution ($0^{_{''}}_{^.}03$) and depth (10,000 s) limits of JWST using ResNet. A U-Net model is employed to pinpoint these small lenses in images, which are otherwise virtually impossible for human detection. Our results indicate that JWST can find ∼17/deg 2 such low-halo-mass lenses, with the locations of ∼1.1/deg 2 of these detectable by the U-Net at ∼100% precision (and ∼7.0/deg 2 at a 99.0% precision). To validate our model for finding “conventional” strong lenses, we apply it to HST images, discovering two new strong lens candidates previously missed by human classifiers in a crowdsourcing project (E. O. Garvin et al. 2022). This study demonstrates the (potentially “superhuman”) advantages of ML combined with current and future space telescopes for detecting conventional, and especially, low-halo-mass strong lenses, which are critical for testing cold dark matter models.
This study explores machine learning for decoding Modbus RTU data using K-Nearest Neighbors (KNN) models. An initial KNN model trained on 8,000 packets achieved 95.15% accuracy. Although ML improves generalization, accuracy still falls short of deterministic methods. These findings have implications for Modbus traffic analysis, intrusion detection in industrial networks, and adaptive error correction in real-time monitoring systems. By refining ML-based decoding, future work could enable more efficient anomaly detection and predictive maintenance in industrial automation and cybersecurity applications.
PBL height from random forest, random forest classifier and lightGBM ML models.