SEARCH · Engineering Papers
Results for “Learning with errors”
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.
Probing Accuracy-Speedup Tradeoff in Machine Learning Surrogates for Molecular Dynamics Simulations
The performance promise of machine learning surrogates of molecular dynamics simulations of soft materials is significant but generally comes at the cost of acquiring large training datasets to learn the complex relationships between input soft material attributes and output properties. Under the constraint of limited high-performance computing resources, optimizing the size of the training datasets becomes paramount. Using an artificial neural network based surrogate for molecular dynamics simulations of confined electrolytes, we explore the tradeoff between surrogate accuracy and computational gains. Accuracy is assessed by computing the root-mean-square errors between the surrogate predictions and the ground truth results obtained via molecular dynamics simulations. The computational performance is judged by evaluating the speedup which incorporates the training dataset creation time. Improvement in accuracy occurs with a loss of speedup, which scales as the inverse of the training dataset size. Furthermore, the link between surrogate generalizability and the accuracy-speedup tradeoff is assessed by examining the errors incurred in surrogate predictions on unseen, interpolated input variables and developing a net speedup metric to capture the associated gains.
Enhancing the formation of ionic defects to study the ice Ih/XI transition with molecular dynamics simulations
Ice Ih, the common form of ice in the biosphere, contains proton disorder. Its proton-ordered counterpart, ice XI, is thermodynamically stable below 72 K. However, the formation of ice XI is kinetically hindered, and experimentally it is obtained by doping with KOH. Doping creates ionic defects that promote the migration of protons and the associated change in proton configuration. In this article, we mimic the effect of doping with a bias potential that enhances the formation of ionic defects in molecular dynamics simulations. The recombination of the ions thus formed proceeds through fast migration of the hydroxide along hydrogen bond loops, providing a physical and expedite way to change the proton configuration. Here, a key ingredient of this approach is a machine learning potential trained with density functional theory data and capable of modelling molecular dissociation. We exemplify the usefulness of this idea by studying the order-disorder transition using an appropriate order parameter that distinguishes the proton environments in ice Ih and XI. We calculate the changes in free energy, enthalpy, and entropy associated with the transition. Our estimated entropy agrees with experiment within the error bars of the calculation.
Symbolic pregression: Discovering physical laws from distorted video
In this work, we present a method for unsupervised learning of equations of motion for objects in raw and optionally distorted unlabeled synthetic video (or, more generally, for discovering and modeling predictable features in time-series data). We first train an autoencoder that maps each video frame into a low-dimensional latent space where the laws of motion are as simple as possible, by minimizing a combination of nonlinearity, acceleration, and prediction error. Differential equations describing the motion are then discovered using Pareto-optimal symbolic regression. We find that our pre-regression (“pregression”) step is able to rediscover Cartesian coordinates of unlabeled moving objects even when the video is distorted by a generalized lens. Using intuition from multidimensional knot theory, we find that the pregression step is facilitated by first adding extra latent space dimensions to avoid topological problems during training and then removing these extra dimensions via principal component analysis. An inertial frame is autodiscovered by minimizing the combined equation complexity for multiple experiments.
Experiences Detecting Defective Hardware in Exascale Supercomputers
In May 2022, the newest supercomputer to top the TOP 500 list was Frontier at Oak Ridge National Laboratory, demonstrating the capability of computing more than 1.1 quintillion (1018) floating-point calculations every second. Driving this ground-breaking rate of computing is Frontier’s more than 37,000 graphics processing units (GPUs) and 9,408 central processing units (CPUs). In total, Frontier contains more than 60 million parts. At this scale, the smallest margin of error may generate hundreds of hardware errors across the system. These errors are capable of directly hindering world-class science performed on Frontier if not found. In this work, we describe and evaluate two strategies for finding hardware-level faults in Frontier’s 9,408 compute nodes. There are two strategies developed: the first uses the Slurm scheduler to scavenge available compute time to run the node screen, the second builds upon the lessons learned in the first strategy and enforces a weekly screen of each node. Using June 2023 as a case study, we find that the first scheduling strategy consumed more than ten times the resources as the second scheduling strategy, but successfully detected five hardware defects in Frontier. We summarize the lessons learned while developing and running a node screen on the world’s first exascale supercomputer.
Learning to Predict and Improve Build Successes in Package Ecosystems
Software has become increasingly complex, with a typical application depending on tens or hundreds of packages. Finding compatible versions and build configurations of these packages is challenging. This paper presents a method to learn the likelihood of software build success, and techniques for leveraging this information to guide dependency solvers to better software configurations. We leverage the heavily parameterized package recipes from the Spack package manager to produce a training data set of builds, and we use Graph Neural Networks to learn whether a given package configuration will build successfully or not. We apply our tool to the U.S. Exascale Computing Project’s software stack. We demonstrate its effectiveness in predicting whether a given package will build successfully. We show that our technique can be used to improve the solutions generated by dependency solvers, reducing the need for developers to find working builds by trial and error.
Blind photovoltaic modeling intercomparison: A multidimensional data analysis and lessons learned
The Photovoltaic (PV) Performance Modeling Collaborative (PVPMC) organized a blind PV performance modeling intercomparison to allow PV modelers to blindly test their models and modeling ability against real system data. Measured weather and irradiance data were provided along with detailed descriptions of PV systems from two locations (Albuquerque, New Mexico, USA, and Roskilde, Denmark). Participants were asked to simulate the plane-of-array irradiance, module temperature, and DC power output from six systems and submit their results to Sandia for processing. The results showed overall median mean bias (i.e., the average error per participant) of 0.6% in annual irradiation and –3.3% in annual energy yield. While most PV performance modeling results seem to exhibit higher precision and accuracy as compared to an earlier blind PV modeling study in 2010, human errors, modeling skills, and derates were found to still cause significant errors in the estimates.
Inverse design of photonic surfaces via multi fidelity ensemble framework and femtosecond laser processing
We demonstrate a multi-fidelity (MF) machine learning ensemble framework for the inverse design of photonic surfaces, trained on a dataset of 11,759 samples that we fabricate using high throughput femtosecond laser processing. The MF ensemble combines an initial low fidelity model for generating design solutions, with a high fidelity model that refines these solutions through local optimization. The combined MF ensemble can generate multiple disparate sets of laser-processing parameters that can each produce the same target input spectral emissivity with high accuracy (root mean squared errors < 2%). SHapley Additive exPlanations analysis shows transparent model interpretability of the complex relationship between laser parameters and spectral emissivity. Finally, the MF ensemble is experimentally validated by fabricating and evaluating photonic surface designs that it generates for improved efficiency energy harvesting devices. Our approach provides a powerful tool for advancing the inverse design of photonic surfaces in energy harvesting applications.
Preliminary Results on Bayesian Inverse UQ for OECD/NEA WPNCS Subgroup 14 Benchmark Exercise for Error Recovery and Experimental Coverage
The Organization for Economic Cooperation and Development (OECD) Working Party on Nucelar Criticality Safety (WPNCS) has proposed a benchmark exercise representative of neutronic behavior in criticality experiments. Here, the goal is to develop confidence in data assimilation techniques used to adjust nuclear data. Participants are given synthetic experimental models with associated measured data and asked to estimate the model parameters given the model and measurements as well as provide predictions for separate application models. In this work, we performed data assimilation using Bayesian inverse Uncertainty Quantification (UQ) with machine learning surrogate models to produce posterior parameter distributions for the requested parameters and posterior predictive distributions for the requested responses. Several experimental models are shown to insufficiently inform the posterior parameter distributions for the applications involved. However, given sufficient experimental data, posterior parameter estimates yielded reduced uncertainty in the response predictions of interest while covering the experimental data.
Invariant Molecular Representations for Heterogeneous Catalysis
Catalyst screening is a critical step in the discovery and development of heterogeneous catalysts, which are vital for a wide range of chemical processes. In recent years, computational catalyst screening, primarily through density functional theory (DFT), has gained significant attention as a method for identifying promising catalysts. However, the computation of adsorption energies for all likely chemical intermediates present in complex surface chemistries is computationally intensive and costly due to the expensive nature of these calculations and the intrinsic idiosyncrasies of the methods or data sets used. This study introduces a novel machine learning (ML) method to learn adsorption energies from multiple DFT functionals by using invariant molecular representations (IMRs). To do this, we first extract molecular fingerprints for the reaction intermediates and later use a Siamese-neural-network-based training strategy to learn invariant molecular representations or the IMR across all available functionals. Our Siamese network-based representations demonstrate superior performance in predicting adsorption energies compared with other molecular representations. Notably, when considering mean absolute values of adsorption energies as 0.43 eV (PBE-D3), 0.46 eV (BEEF-vdW), 0.81 eV (RPBE), and 0.37 eV (scan+rVV10), our IMR method has achieved the lowest mean absolute errors (MAEs) of 0.18 0.10, 0.16, and 0.18 eV, respectively. These results emphasize the superior predictive capacity of our Siamese network-based representations. The empirical findings in this study illuminate the efficacy, robustness, and dependability of our proposed ML paradigm in predicting adsorption energies, specifically for propane dehydrogenation on a platinum catalyst surface.
Utilizing machine learning to predict tensile ductility and yield strength of CoNiV-based multi-principal elements alloys
This study explores the use of machine learning (ML) as a computational tool to accelerate the design of multi-principal element alloys (MPEAs) with improved tensile elongation. An ML model was trained using available experimental data from the literature along with theoretically derived features to predict yield strength (YS) and ductility. A subset of ML-predicted compositions—CoNiVFe, CoNiVTi, CoNiVTiFe, and CoCrNiVTi—was synthesized and evaluated through tensile testing. The ML model underpredicted YS by approximately 20–30 % and overpredicted ductility by 60–70 % for Ti-containing alloys. Microstructural analysis revealed that Ti segregation at interdendritic regions contributed to early fracture, leading to discrepancies in ductility predictions. Ti segregation at these regions likely drives the increased YS due to segregation strengthening. In contrast, the CoNiVFe alloy showed good agreement with both experimental YS and elongation, with prediction errors of ∼10.2 % and ∼20.7 %, respectively. Microstructural characterization revealed minimal segregation in this alloy, suggesting that the ML model can reliably predict the properties of alloys with little to no segregation. These findings highlight the capability of ML in predicting YS with good accuracy but underscore its limitations in capturing defect-driven failure mechanisms such as segregation-induced embrittlement.
SMC 2021 : Analyzing Resource Utilization and User Behavior on Titan Supercomputer
Resource utilization statistics of submitted jobs on a supercomputer can help us understand how users from various scientific domains use HPC platforms and better design a job scheduler. We explore to generate insight regarding workload distribution and usage pattern domains from job scheduler trace, GPU failure information, and project-specific information collected from Titan supercomputer. Furthermore, we want to know how the scheduler performance varies over time and how the users' scheduling behavior changes following a system failure. These observations have the potential to provide valuable insight, which is helpful to prepare for system failures. These practices will help us develop and apply novel machine learning algorithms in understanding system behavior, requirement, and better scheduling of HPC systems. There are two datasets, RUR and GPU. RUR: This dataset is the job scheduler traces collected from the Titan supercomputerfrom 01/01/2015 to 07/31/2019 (2015.csv - 2019.csv). These were collected usingResource Utilization Report (RUR), a Cray-developed resource-usage data collectionand reporting system. It contains the usage information of its critical resources (CPU,Memory, GPU, and I/O) of each running job on Titan during that period [2]. ProjectAreas: Every job is associated with a project ID. TheProjectAreas.csvdatasetprovides a mapping of the project ID to its domain science. GPU: There have been some hardware-related issues in the GPUs in Titan that caused some GPUs to fail, sometimes irrecoverably during some job runs. This dataset provides information regarding these failures during the execution of the submitted jobs. GPUs on Titan are uniquely identified by a serial number (SN), and they are installed in a location. A GPU can be installed in a location, then removed from that location following a failure, and then re-installed in a different location after fixing the problem. If the failure can't be recovered, the GPU might be removed entirely from Titan. There are two prominent types of failures that resulted in the removal of GPUs from Titan: Double Bit Error (DBE) and Out of the Bus (OTB). The dataset (gc_full.csv) has the following fields: 1. SN : Serial number of a GPU 2. location : The location where it is installed 3. insert : The time when it was inserted into that location 4. remove : The time when it was removed from that location 5. duration : Amount of time the GPU spent in this location 6. out : If the device was taken out entirely w/o a re-installment into a new location. 7. event : If the GPU was taken out entirely, the reason for its removal. To learn more about this dataset, please refer to the git repositoryhttps://github.com/olcf/TitanGPULifeand the related publication [1]. References [1] George Ostrouchov, Don Maxwell, Rizwan A Ashraf, Christian Engelmann, MallikarjunShankar, and James H Rogers. Gpu lifetimes on titan supercomputer: Survival analysisand reliability. InSC20: International Conference for High Performance Computing,Networking, Storage and Analysis, pages 1-14. IEEE, 2020. [2] Feiyi Wang, Sarp Oral, Satyabrata Sen, and Neena Imam. Learning from five-yearresource-utilization data of titan system. In2019 IEEE International Conference onCluster Computing (CLUSTER), pages 1-6. IEEE, 2019.
Data-driven acoustic measurement of moisture content in flowing biomass
Measuring the moisture content in flowing biomass is critical to processes such as liquid biofuel conversion, such as biogasoline, biodiesel, bio jet kerosene, etc. However, biomass tends to flow in aggregates, which results in significant inhomogeneities in the amount of biomass flowing in front of a sensor at a given time, and there can be significant overlap in the material properties of dry vs wet biomass, leading to poor signal-to-noise ratio. We present a technique for identifying biomass moisture content using a series of acoustic pitch-catch measurements to quantify the sound speed and acoustic amplitude through the biomass, in conjunction with classical machine learning techniques, including Naive Bayes, Random Forest, and K-Nearest Neighbors classification. We amplify the differences between the acoustic measurements in different moisture levels by collecting a series of pulse-echo measurements, which we sort in order of ascending sound speed. We test the accuracy of the technique on experimentally-prepared batches of corn stover biomass with specified moisture levels and measure the average error in the estimated moisture level as a function of the number of pitch-catch measurements used. We observe average estimation errors as low as 6.7% by increasing the number of measurements and optimizing the hyperparameters. This work presents a novel method determining moisture content in flowing biomass with inhomogeneous flow. Additionally, this technique has application in optimizing biomass conversion processes, as well as other fields including, paper production, natural fiber processing, and mineral extraction.
Machine learning surrogates for ion energy–angle distributions in thermal and RF plasma sheaths
Ion energy–angle distributions (IEADs) at material surfaces are a critical input for plasma–material interaction (PMI) studies in fusion devices, yet they are computationally expensive to obtain using particle-in-cell (PIC) simulations. In this work, we develop a machine learning surrogate based on a deep deconvolutional neural network (DDeCNN) trained on large databases generated with the hPIC2 code. The surrogate is capable of reconstructing IEADs from sheath parameters for both thermal and radio-frequency (RF) plasmas, including cases with multiple ion species. Across thousands of test cases, the model achieves high accuracy, with over 97 % of predictions classified as good or average based on standard error metrics (MAE, MSE, L2). Even in the more challenging RF and multi-species regimes, the surrogate reliably captures the multi-peak structure of PIC results. Once trained, the surrogate produces IEADs in milliseconds on a common workstation, yielding speedups of six to seven orders of magnitude compared with running a full PIC simulation. This computational gain enables dense parameter scans and direct coupling of IEAD predictions with PMI and erosion models on whole-device scales in fusion-relevant conditions.
Conversion of METL Diagnostic Flowmeters to Permanent Magnet Based Electromagnetic Flowmeters
The Mechanisms Engineering Test Loop (METL) uses three flowmeters to monitor sodium circulation through the main loop, the Cold Trap leg, and the Plugging Meter leg. When constructed, the Cold Trap flowmeter (CTFM) and Plugging Meter flowmeter (PMFM) were equipped with permanent magnets and Hall Sensors. By October of 2019, during normal operations, both the CTFM and PMFM failed, producing either erroneous or zero signals. Repairs to the flowmeters were attempted in January of 2020 but failed to resolve the issues. Therefore, starting in March of 2021, a plan was developed to convert the installed flowmeters to more robust permanent magnet based electromagnetic flowmeters (EMFMs). The Cold Trap EMFM (CT EMFM) and Plugging Meter EMFM (PM EMFM) were fully installed by August of 2021. Initial testing and commissioning were completed in November of 2021. Finally, the first calibration runs were finished in March of 2022. This report discusses the design, conversion, and calibration of the as-built CTFM and PMFM to robust permanent magnet based EMFMs. During the design phase, theoretical models were used to estimate the signal output and required magnet sizes. Then, the CTFM and PMFM were disassembled and converted to EMFMs. After commissioning, the CT EMFM and PM EMFM were calibrated using a ‘Time Transfer Procedure’ carried out with METL Expansion Vessels and Test Vessel Two. The accuracy of the calibration was estimated through an uncertainty analysis. Finally, the report describes some lessons learned for future flowmeter conversions and follow-up work. In total, both METL EMFM’s met the desired performance targets. Testing data showed that the METL EMFMs provide a output signal of roughly 2 mV at 1 GPM and 250 °C. During calibration, flowrates between 0.7 GPM and 2.4 GPM were achieved. In this flow range, calibration coefficients of 0.485 and 0.935 were measured as with errors ranging from 15-5% and 16-5% for the CT EMFM and PM EMFM, respectively. Above 1.5 GPM, the error was assumed to be about 5% for both METL EMFMs. Several points will be considered for future follow-up work on the METL EMFM’s. First, the signal noise issues in the PM EMFM will be addressed. It is likely the source of the noise is due to some stray EMF from a heater. Secondly, the NdFeB magnets will be replaced with a grade that has a Curie temperature above 80 °C. For example, a few grades of NeFeB magnets have Curie temperatures which exceed 150 °C and some SmCo magnets have Curie temperatures as high as 500 °C. Third, more calibration runs will be completed in the range of 0.5-2.5 GPM. This will further improve the accuracy of the calibration coefficient. Fourth, calibration runs will be completed up to 5 GPM. This requires a larger pressure relief valve to be installed on the Expansion Vessel and Test Vessel 2, as well as a revised calibration procedure. Lastly, calibration runs will be completed to account for cold spots in the system, which reduced the inlet temperature of the METL EMFMs. Overall, these points will improve the calibration coefficients derived in this work and also improve the performance of the METL EMFMs.
A Comparison of Infectious Disease Forecasting Methods across Locations, Diseases, and Time
Accurate infectious disease forecasting can inform efforts to prevent outbreaks and mitigate adverse impacts. This study compares the performance of statistical, machine learning (ML), and deep learning (DL) approaches in forecasting infectious disease incidences across different countries and time intervals. We forecasted three diverse diseases: campylobacteriosis, typhoid, and Q-fever, using a wide variety of features (n = 46) from public datasets, e.g., landscape, climate, and socioeconomic factors. We compared autoregressive statistical models to two tree-based ML models (extreme gradient boosted trees [XGB] and random forest [RF]) and two DL models (multi-layer perceptron and encoder–decoder model). The disease models were trained on data from seven different countries at the region-level between 2009–2017. Forecasting performance of all models was assessed using mean absolute error, root mean square error, and Poisson deviance across Australia, Israel, and the United States for the months of January through August of 2018. The overall model results were compared across diseases as well as various data splits, including country, regions with highest and lowest cases, and the forecasted months out (i.e., nowcasting, short-term, and long-term forecasting). Overall, the XGB models performed the best for all diseases and, in general, tree-based ML models performed the best when looking at data splits. There were a few instances where the statistical or DL models had minutely smaller error metrics for specific subsets of typhoid, which is a disease with very low case counts. Feature importance per disease was measured by using four tree-based ML models (i.e., XGB and RF with and without region name as a feature). The most important feature groups included previous case counts, region name, population counts and density, mortality causes of neonatal to under 5 years of age, sanitation factors, and elevation. This study demonstrates the power of ML approaches to incorporate a wide range of factors to forecast various diseases, regardless of location, more accurately than traditional statistical approaches.
Constructing a spatiotemporally coherent long-term PM2.5 concentration dataset over China during 1980–2019 using a machine learning approach
The lack of long-term observations and satellite retrievals of health-damaging fine particulate matter in China has demanded the estimates of historical PM 2.5 (particulate matter less than 2.5 µm in diameter) concentrations. This study constructs a gridded near-surface PM 2.5 concentration dataset across China covering 1980–2019 using the space-time random forest model with atmospheric visibility observations and other auxiliary data. The modeled daily PM2.5 concentrations are in excellent agreement with ground measurements during 2015–2019, with a coefficient of determination of 0.95 and mean relative error of 12%. Besides the atmospheric visibility which explains 30% of total importance of variables in the model, both emissions and meteorological conditions are also essential factors affecting PM2.5 predictions. From 1980 to 2014, the model-predicted PM2.5 concentrations increased constantly with the maximum growth rate of 5–10 µg/m3/decade over eastern China. Due to the clean air actions, PM 2.5 concentrations have decreased effectively at a rate over 50 µg/m 3 /decade in the North China Plain and 20–50 µg/m 3 /decade over eastern China during 2014–2019. The newly generated dataset of 1-degree gridded PM 2.5 concentrations for the past 40 years across China provides a useful means for investigating interannual and decadal environmental and climate impacts related to aerosols.
Streamlining Ocean Dynamics Modeling with Fourier Neural Operators: A Multiobjective Hyperparameter and Architecture Optimization Approach
Training an effective deep learning model to learn ocean processes involves careful choices of various hyperparameters. We leverage DeepHyper’s advanced search algorithms for multiobjective optimization, streamlining the development of neural networks tailored for ocean modeling. The focus is on optimizing Fourier neural operators (FNOs), a data-driven model capable of simulating complex ocean behaviors. Selecting the correct model and tuning the hyperparameters are challenging tasks, requiring much effort to ensure model accuracy. DeepHyper allows efficient exploration of hyperparameters associated with data preprocessing, FNO architecture-related hyperparameters, and various model training strategies. We aim to obtain an optimal set of hyperparameters leading to the most performant model. Moreover, on top of the commonly used mean squared error for model training, we propose adopting the negative anomaly correlation coefficient as the additional loss term to improve model performance and investigate the potential trade-off between the two terms. The numerical experiments show that the optimal set of hyperparameters enhanced model performance in single timestepping forecasting and greatly exceeded the baseline configuration in the autoregressive rollout for long-horizon forecasting up to 30 days. Utilizing DeepHyper, we demonstrate an approach to enhance the use of FNO in ocean dynamics forecasting, offering a scalable solution with improved precision.