Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model error”

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 1,747 records · Page 97

MO-D-BRD-02: In Memoriam of Bengt Bjarngard: SBRT II: Small Field Dosimetry - TG155

Specialized radiation treatment such as SRS/SRT. SBRT, IMRT, VMAT, Tomotherapy, CyberKnife and Gamma Knife use small fields or combination of small fields where dosimetry is challenging and uncertain due to non-equilibrium conditions such as longitudinal and lateral disequilibrium. Additionally the primary photon fluence is greatly affected by the obstruction of the source size by the jaws creating a large dose gradient across the field. Electronic equilibrium is a phenomenon associated with the range of secondary particles which depend on the beam energy, photon spectrum and the composition of the medium. Additionally, the finite size of detectors creates volume averaging and fluence perturbations especially in small fields. The IAEA/AAPM has provided a frame work for non-compliant reference dosimetry in small fields1. The AAPM TG-1552 has adopted this frame work to provide guidelines in relative dosimetry. This course provides the insight of TG-155 that defines small field, provides recommendations for suitable detectors and associated correction factors to convert reading to dose. Recommendations of a good working practice for relative dosimetry measurements (PDD, TMR, output factor, etc.) and dose calculations based on the new formulation is are elaborated. It also discusses beam modeling and dose calculations as a critical step in clinical utilization of small field radiotherapy. Small errors in beam data, approximations in dose algorithms, or misaligned of detectors and field settings can propagate into large errors in planned and delivered dose. The modeling and treatment planning aspects of small field dosimetry are reviewed with emphasis on the most critical parts for ensuring accurate and safe radiation therapy. Discussion on k(fmsr, fclin) for commercially available detectors are also provided.1 P. Alfonso, P. Andreo, R. Capote, M. S. Huq, W. Kilby, P. Kjall, T. R. Mackie, H. Palmans, K. Rosser, J. Seuntjens, W. Ullrich and S. Vatnitsky, “A new formalism for reference dosimetry of small and nonstandard fields,” Med Phys 35, 5179–5186 (2008).2 I. J. Das, P. Francescon, A. Ahnesjo, M. M. Aspradakis, C. W. Cheng, G. X. Ding, G. S. Ibbott, M. Oldham, M. S. Huq, C. S. Reft and S. O.A., “Task Group 155 report: Small fields and non-equilibrium condition photon beam dosimetry,” Med Phys (under review). Learning Objectives: Physics of cavity theory for non-equilibrium condition and definition of small fields Description of absolute and relative dosimetry in the context of TG-155 Choice of detector in small field dosimetry, perturbations and corrections (K(fmsr, fclin)) for accurate dosimetry Understand the advantages and limitations of commercially available detectors for small field dosimetry.

60 APPLIED LIFE SCIENCES↗

Fire Risk Investigation in 3D (FRI3D) Software and Process for Integrated Fire Modeling

Modeling and implementing fire safety for nuclear power plants is a costly activity. Because of the complexity of fire phenomena and multiple operational procedures, it is difficult to computationally provide assurance that mitigation methods are adequate for critical areas using current analysis methods in a cost-effective manner. The nuclear industry needs efficient methods to provide more accurate modeling and optimize mitigation methods to improve nuclear power plant understanding for fire scenarios. This report describes the work done over the last two year on the Fire Risk Investigation in 3D (FRI3D) software. This software simplifies the modeling process by importing existing plant models and data, incorporating a 3D modeling environment and coupling with the fire simulation software CFAST. This eliminates the manual labor and errors associated with much of the fire modeling process. The report lays out the status and path of FRID software for the Risk Informed Safety Analysis pathway of the Light Water Reactor Sustainability project at Idaho National Laboratory and industry collaborators to best help reduce costs and improve realism in fire probabilistic risk analysis.

97 MATHEMATICS AND COMPUTING↗

Six Machine-Learning Methods for Predicting Hospital-Stay Duration for Patients with Sepsis: A Comparative Study

Sepsis is a life-threatening medical condition that, if not treated promptly, can result in tissue damage, organ failure, and death. According to the Centers for Disease Control, about 270,000 individuals die of sepsis in the US each year. Further, sepsis expenditures accounted for 13% of total US hospital costs in 2013, totaling more than $24 billion. Our project objectives were to determine if Machine Learning algorithms could reliably predict hospital stay duration for patients with sepsis. The data set we used has been de-identified and is freely available through the BupaR package. The data includes 1050 cases, 15214 events, and 16 types of actions related to sepsis patient care. First, we used process mining to determine how long each patient was in the hospital. Using BupaR’s functions, we created several process model graphs. These process models depict the movement of patients at a hospital and provide duration data for each patent case. Second, we identified outlier data and created two dataset versions: one with and one without outliers. We then applied the following analysis methods: Linear Regression, Random Forest, K-Nearest Neighbors, Neural Networks, XGBoost, and lightGBM. We compared the model validations for the six machine learning models using the same data-splitting method. We found that the XGBoost model had the best prediction accuracy of 73.9 percent for cases with outliers, and 79 percent for cases without outliers. We also found that the lightGBM model had the lowest mean absolute error between prediction and actual duration in days with 3.66 days for the case with outliers, and 2.4 days for the case without outliers. These two models outperformed the other four models. This work will be enhanced in the future by exploring new prediction algorithms and comparing them with the results of this study.

Chen, Lingtao↗

Linear system identification of a cold flow circulating fluidized bed

Knowledge of the solids circulation rate (SCR) is essential to the control and improved performance of a circulating fluidized bed system. In the present work, the noise model is derived using the prediction error method considering process and measurement noises acting on the cold flow circulating fluidized bed (CFCFB) with a cork particulate material. The outputs of the initial model are the total pressure drop across the riser, the pressure drop across the crossover, the pressure drop across the primary cyclone, the total pressure drop across the stand-pipe, the pressure drop across the loop seal, and the SCR. The stochastic estimate of SCR is determined from the noise model using the stochastic pressure drop estimates. The deterministic estimate is obtained through the inputs taken as move air flow, riser aeration, and loop seal fluidization air that are all independent variables of the given setup and under the control of the user. The theory has been developed to convert a complete blackbox model to a grey box model through the output-to-state transformation such that both the models of the CFCFB consists of all these output variables as the states of the system, and only pressure drops across the system as the output measurements. Thus, the final models do not include any fictitious terms and they are defined only in terms of physical parameters of the given system. Both components of SCR are separately analysed. The combined SCR response of both the noise model and deterministic model is compared with the validation data set of this state variable in terms of modelfit, and the results are shown.

42 ENGINEERING↗

Predicting Elastic Constants of Refractory Complex Concentrated Alloys Using Machine Learning Approach

Refractory complex concentrated alloys (RCCAs) have drawn increasing attention recently owing to their balanced mechanical properties, including excellent creep resistance, ductility, and oxidation resistance. The mechanical and thermal properties of RCCAs are directly linked with the elastic constants. However, it is time consuming and expensive to obtain the elastic constants of RCCAs with conventional trial-and-error experiments. The elastic constants of RCCAs are predicted using a combination of density functional theory simulation data and machine learning (ML) algorithms in this study. The elastic constants of several RCCAs are predicted using the random forest regressor, gradient boosting regressor (GBR), and XGBoost regression models. Based on performance metrics R-squared, mean average error and root mean square error, the GBR model was found to be most promising in predicting the elastic constant of RCCAs among the three ML models. Additionally, GBR model accuracy was verified using the other four RHEAs dataset which was never seen by the GBR model, and reasonable agreements between ML prediction and available results were found. The present findings show that the GBR model can be used to predict the elastic constant of new RHEAs more accurately without performing any expensive computational and experimental work.

36 MATERIALS SCIENCE↗

Transplatformer: translating toxicogenomic profiles between generations of platforms

Background Transcriptomic profiling technologies have advanced the analysis of biological and toxicological responses. However, substantial differences in probe design, dynamic range, gene coverage, and preprocessing pipelines across platforms introduce artifacts that limit cross-study integration and hinder the reuse of historical datasets. We aim to develop computational methods for accurate cross-platform translation to maximize the value of legacy resources. Results We present TransPlatformer a deep learning framework for translating gene expression profiles across heterogeneous toxicogenomics platforms. TransPlatformer employs a novel attention-based architecture to map high-dimensional fold-change vectors from legacy microarray technologies to current platforms. Models are trained and evaluated using DrugMatrix, spanning three technological generations. We investigate mixed-tissue, single-tissue, and cross-tissue training paradigms and benchmark performance against multilayer perceptron and matrix-completion baselines. In mixed-tissue training, TransPlatformer achieves a greater than 50% reduction in mean absolute error (0.043 vs. 0.09) and nearly doubles Pearson correlation ( ≈ 0.71 vs. 0.37) relative to baseline methods. Importantly, TransPlatformer preserves rare but biologically meaningful over- and under-expressed signals, with mean absolute error below 0.22. Single-tissue models yield further improvements for well-represented organs, such as a 10% reduction in liver mean absolute error, while underscoring the need for data augmentation strategies in low-sample tissues.ra Conclusions TransPlatformer provides an effective and scalable computational solution for cross-platform transcriptomic translation. By enabling biologically faithful harmonization of gene expression data, the proposed approach facilitates the reuse of legacy toxicogenomics datasets, enhances downstream biomarker discovery, and supports more reproducible predictive modeling in toxicology.

59 BASIC BIOLOGICAL SCIENCES↗

An Assessment of the Error Due to Computing Waste Isolation Pilot Plant Porosity Using the Porosity Response Surface Approach

The Waste Isolation Pilot Plant Performance Assessment (WIPP PA) must predict the likelihood that radionuclides will escape into the biosphere via mechanisms that depend on geohydraulic flow. Ideally, one would predict the geohydraulic flow using coupled geohydraulic and geomechanical simulations, but such coupled simulations are not computationally tractable. Instead, Sandia has historically used a look-up table of porosities for a given fluid pressure and time, called the porosity response surface, but this approach can introduce porosity errors because it largely ignores the porosity’s dependence on the past fluid pressure history. This report discusses efforts to quantify these porosity errors for both the legacy and new porosity response surfaces. Six hundred different fluid pressure histories were fed through the legacy/new geomechanical model and the legacy/new porosity response surface to generate six hundred porosity error histories. The error associated with the legacy porosity surface was substantial, while the error associated with the new porosity surface was typically small, except when fluid pressures exceeded the lithostatic pressure at the repository. In response to the errors at high pressures, a preliminary study of the WIPP PA’s sensitivity to these porosity errors was conducted. The study found that reducing the porosity errors at high pressures negligibly affected predictions of radionuclide releases. Finally, an initial machine-learned model for porosity was developed. This ML model significantly reduced the porosity error at high pressures, but sizable errors remained, so more development is necessary before coupling an ML model to the geohydraulic model.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Toward more-robust, AI-enabled subsurface seismic imaging for geotechnical applications

Non-invasive seismic imaging has the potential to cost-effectively evaluate large volumes of subsurface material to inform geotechnical site investigation. However, seismic imaging using full waveform inversion (FWI) requires significant computational time and is dependent on an initial starting model. As a result, FWI has not yet been widely adopted into geotechnical practice. Previous efforts, on relatively simple two-layered models, indicate that data-driven artificial intelligence (AI) models may be as effective as FWI at predicting 2D images of shear wave velocity (V s ). Furthermore, the AI model predictions can be made almost instantaneously after data acquisition and do not require an initial starting model. We examine the generality of these findings by developing a new AI model for subsurface seismic imaging, whereby we make several notable contributions. First, we architect a multimodal AI model that combines time- and frequency-domain representations of the seismic wavefield to predict a 50 m by 20 m subsurface image of V s . Second, we developed a new diverse dataset of 100,000 images with their corresponding seismic wavefields to train the AI model. Third, we propose four physics-informed data augmentations for data-driven seismic imaging. Fourth, we develop two prediction consistency tests to evaluate the model’s performance when the true subsurface is unknown. Our final model, which has been made publicly available, is capable of predicting a subsurface V s image from a single seismic wavefield with an average, mean absolute percent error (MAPE) of 24 %. The predictive model is applied to a field dataset and shown to be consistent with local geology and shear-wave refraction measurements from the same location.

Artificial intelligence↗

Influence of the line-shape model on the spectroscopic determination of the Boltzmann constant

Recent high-resolution spectroscopic experiments demonstrated the possibility of measuring the Boltzmann constant directly from the Doppler width of atomic or molecular lines; however, both experimental setups and data analysis methods still need to be improved to compete with the best acoustic methods. In this paper, the influence of choice of spectral line-shape model for data analysis on systematical error of the Doppler-width determination is analyzed on the example of the oxygen line near {lambda}=687 nm. Profile simulations were performed for many different models at pressures ranging from 0.012 Pa to 120 kPa, taking into account such line-shape effects as the speed dependence of collisional broadening and Dicke narrowing. We discuss limitations of an applicability of the Voigt profile for purposes of the Boltzmann constant determination. The influence of the model line-shape profile and the gas pressure range on the Doppler width extrapolated to the zero-pressure value is investigated.

74 ATOMIC AND MOLECULAR PHYSICS↗

Longwave 3D Benchmarks for Inhomogeneous Clouds and Comparisons with Approximate Methods

The purpose of this study is twofold: to (i) establish three-dimensional (3D) longwave radiative transfer benchmarks for inhomogeneous cloud fields and (ii) compare the results with three approximate, 1D methods. The benchmark results are calculated using a correlated-k three-dimensional Monte Carlo (3DMC) algorithm that is validated via comparisons to line-by-line calculations for simple atmospheres. The approximate methods include an independent column approximation (ICA) and two cloud-overlap schemes: maximum/random (MRO) and random (RO). Six inhomogeneous cloudy-sky test cases are used and encompass a wide range of domain sizes used by general circulation models. Domain-averaged fluxes and heating rates from these atmospheres show that the ICA is consistently more accurate than the cloud-overlap models with respect to the 3D benchmarks. For example, comparisons of model results for the Atlantic Trade Wind Experiment (ATEX), a marine boundary layer cumulus field, yield a maximum cloud-layer heating rate error of 15.73 K day -1 from using cloud-overlap models, whereas the ICA error is only 2.17 K day -1 . This paper presents results showing that these differences are attributed to the 3D effects of unresolved clouds and indicate that there is an inherent deficiency in the ability of 1D models to accurately calculate radiative quantities in these atmospheres.

54 ENVIRONMENTAL SCIENCES↗

Electron-Proton Scattering Event Generation using Structured Tokenization

Recent work such as Omnijet-$\alpha$ has demonstrated that effective tokenization combined with transformer-based architectures can produce effective foundation models for jet physics. While tokenization may help models capture generalizable event characteristics, it also introduces discretization errors that may compromise the precision required for downstream physics analyses. As the number and complexity of the particle features grow, these errors are likely to grow proportionally. In this study, we investigate new tokenization strategies to improve the application of generative transformer models to \textsc{Pythia8} simulations of electron-proton scattering at the Electron-Ion Collider. Specifically, we propose a feature-based structured tokenization approach that utilizes multiple tokens per particle, improving expressivity, while reducing the total number of unique tokens needed. We evaluate this method against grid-based binning, K-means clustering, and vector-quantized variational auto-encoders on the event simulations. Our results show that feature-based structured tokenization reduces discretization error, leading to more accurate generative modeling of particle-level events.

Goldenberg, Steven [Thomas Jefferson National Acce↗

Electron-Proton Scattering Event Generation using Structured Tokenization

Recent work such as Omnijet-$\alpha$ has demonstrated that effective tokenization combined with transformer-based architectures can produce effective foundation models for jet physics. While tokenization may help models capture generalizable event characteristics, it also introduces discretization errors that may compromise the precision required for downstream physics analyses. As the number and complexity of the particle features grow, these errors are likely to grow proportionally. In this study, we investigate new tokenization strategies to improve the application of generative transformer models to \textsc{Pythia8} simulations of electron-proton scattering at the Electron-Ion Collider. Specifically, we propose a feature-based structured tokenization approach that utilizes multiple tokens per particle, improving expressivity, while reducing the total number of unique tokens needed. We evaluate this method against grid-based binning, K-means clustering, and vector-quantized variational auto-encoders on the event simulations. Our results show that feature-based structured tokenization reduces discretization error, leading to more accurate generative modeling of particle-level events.

Goldenberg, Steven [Thomas Jefferson National Acce↗

Cryogenic-Refined MOSFET Modeling for Oscillator, Frequency Divider, and Amplifier Designs Below 4 K

Capturing device characteristic changes at cryogenic temperatures is crucial for cryo-CMOS circuit designs. In this work, we present an isothermal cryogenic-refined modeling approach for CMOS transistors that is simple, low overhead, and easy to implement while offering the required accuracy for predicting circuit performance at the designated temperatures. Guided by die-level measurement data and circuit design principles, the model introduces corrections to only five critical parameters: threshold voltage, carrier mobility, elevated low-frequency flicker noise, dominant high-frequency shot noise, and subthreshold swing (SS). These refinements are implemented around the foundry-provided SPICE model, which is typically validated only down to about 200 K. With these adjustments, the proposed cryogenic-refined model achieves less than 5% error in both large-signal metrics (I–V characteristics) and small-signal parameters (e.g., transconductance) when compared with device measurements at deep-cryogenic temperatures. The methodology is validated in two advanced technologies: TSMC 40-nm CMOS and GlobalFoundries (GF) 45-nm RF-SOI. We further demonstrate its applicability in three representative RF circuits: a 30-GHz LC oscillator, a high-speed current-mode-logic (CML) frequency divider (FD), and a subthreshold Gb/s amplifier, all showing close agreement between simulated predictions and measurements performed at 4 and 2.5 K. Finally, we believe that the proposed approach is implementation-friendly and can significantly accelerate the development of cryo-CMOS integrated circuits.

circuit modeling↗

Bayesian Neural Networks for Uncertainty Analysis of Hydrologic Modeling: A Comparison of Two Schemes

Bayesian Neural Networks (BNNs) have been shown as useful tools to analyze modeling uncertainty of Neural Networks (NNs). This research focuses on the comparison of two BNNs. The first BNNs (BNN-I) use statistical methods to describe the characteristics of different uncertainty sources (input, parameter, and model structure) and integrate these uncertainties into a Markov Chain Monte Carlo (MCMC) framework to estimate total uncertainty. The second BNNs (BNN-II) lump all uncertainties into a single error term (i.e. the residual between model prediction and measurement). In this study, we propose a simple BNN-II, which use Genetic Algorithms (GA) and Bayesian Model Averaging (BMA) to calibrate Neural Networks with different structures (number of hidden units) and combine the predictions from different NNs to derive predictions and uncertainty analysis. We tested these two BNNs in two watersheds for daily and monthly hydrologic simulation. The BMA based BNNs developed in this study outperforms BNN-I in the two watersheds in terms of both accurate prediction and uncertainty estimation. These results show that, given incomplete understanding of the characteristics associated with each uncertainty source, the simple lumped error approach may yield better prediction and uncertainty estimation.

99 GENERAL AND MISCELLANEOUS↗

AGR-5/6/7 Thermal Model with Non-uniform Gas Gaps

Fuel compact temperatures are a crucial factor in assessing irradiation performance of the Tristructural isotropic fuel particles. In the absence of direct measurement, fuel compact temperatures were calculated using a three-dimensional finite element thermal model, which is subject to simulation uncertainty. The most dominant factor in uncertainty of calculated fuel temperature is the gas gap uncertainty due to the nub-to-shell clearance because of fabrication error. A revision to the thermal model was made to examine the most probable offset position for four different dates of interest. The offset position options varied in magnitude and azimuthal direction of offset of the holder top and bottom position. The best-fit offset of the holder is estimated based on the minimum root mean square error of residuals for all operational thermocouples (TCs). From these results, the following conclusions were made: 1) During earlier cycles (162A – 164A) when numerous TCs were still operational, the best-fit offset distance varied in range [0.002 – 0.0035 in.] for both the top and bottom holder while offset azimuthal direction varied widely, especially for the holder bottom. 2) Holder offset led to slightly lower Capsule 1 average temperature, but wider temperature variation (lower minimum and higher peak fuel temperatures). 3) The maximum offset of 0.006 in. to the northwest direction for both the top and bottom holder yielded a much higher temperature than the original peak fuel temperature, which increased from 1422 to 1557 ? (or 135 ? increase). This could be one of the reasons for massive particle failure near the end of Cycle 165A. 4) Even though, the highest temperature at the TC-1-7 tip reached slightly higher than 1000?, the temperature along the TC wire reached as high as 1335? assuming a holder offset of 0.006 in. to the northwest. Consequently, at this temperature nickel from this TC wire could cause particle failures in Capsule 1.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Hybrid electric buses fuel consumption prediction based on real-world driving data

Estimating fuel consumption by hybrid diesel buses is challenging due to its diversified operations and driving cycles. Here, long-term transit bus monitoring data were utilized to empirically compare fuel consumption of diesel and hybrid buses under various driving conditions. Artificial neural network (ANN) based high-fidelity microscopic (1 Hz) and mesoscopic (5–60 min) fuel consumption models were developed for hybrid buses. The microscopic model contained 1 Hz driving, grade, and environment variables. The mesoscopic model aggregated 1 Hz data into 5 to 60-minute traffic pattern factors and predicted average fuel consumption over its duration. The prediction results show mean absolute percentage errors of 1–2% for microscopic models and 5–8% for mesoscopic models. The data were partitioned by different driving speeds, vehicle engine demand, and road grade to investigate their impacts on prediction performance.

33 ADVANCED PROPULSION SYSTEMS↗

Dependence of the bit error rate on the signal power and length of a single-channel coherent single-span communication line (100 Gbit s{sup -1}) with polarisation division multiplexing

We have studied experimentally and using numerical simulations and a phenomenological analytical model the dependences of the bit error rate (BER) on the signal power and length of a coherent single-span communication line with transponders employing polarisation division multiplexing and four-level phase modulation (100 Gbit s{sup -1} DP-QPSK format). In comparing the data of the experiment, numerical simulations and theoretical analysis, we have found two optimal powers: the power at which the BER is minimal and the power at which the fade margin in the line is maximal. We have derived and analysed the dependences of the BER on the optical signal power at the fibre line input and the dependence of the admissible input signal power range for implementation of the communication lines with a length from 30 – 50 km up to a maximum length of 250 km. (optical transmission of information)

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

SubKit

Generating a CTF (thermal-hydraulics subchannel code) model can be time consuming and error prone because of the rigid formatting of the input file and the amount of input requirements. Because of the structured input format, making small changes to a model can require large scale changes to the input file. Furthermore, a user typically has to perform calculations to develop CTF input parameters from actual model geometry. Changing model geometry requires re-calculating input parameters and copying them from wherever the calculation was performed, be it a script, spreadsheet, or piece of paper. This process complicates the quality assurance process for that model and makes changing the model more difficult. CtfDeckBuilder is an object-oriented Python interface to CTF that allows the user to build a CTF model in Python. The user no longer needs to interface with the CTF input file when using the CtfDeckBuilder. The interface reduces the information required of the user by offering defaults for most modeling options provided by CTF. The user can perform calculations directly in Python when building the model so that calculations stay with the model itself.

Salko, Jr., RobertK↗