Search NASA⌕ Search

SEARCH · Search NASA

Results for “optimal accuracy”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Accuracy optimized neural networks do not effectively model optic flow tuning in brain area MSTd

Accuracy-optimized convolutional neural networks (CNNs) have emerged as highly effective models at predicting neural responses in brain areas along the primate ventral stream, but it is largely unknown whether they effectively model neurons in the complementary primate dorsal stream. We explored how well CNNs model the optic flow tuning properties of neurons in dorsal area MSTd and we compared our results with the Non-Negative Matrix Factorization (NNMF) model, which successfully models many tuning properties of MSTd neurons. To better understand the role of computational properties in the NNMF model that give rise to optic flow tuning that resembles that of MSTd neurons, we created additional CNN model variants that implement key NNMF constraints – non-negative weights and sparse coding of optic flow. While the CNNs and NNMF models both accurately estimate the observer's self-motion from purely translational or rotational optic flow, NNMF and the CNNs with nonnegative weights yield substantially less accurate estimates than the other CNNs when tested on more complex optic flow that combines observer translation and rotation. Despite its poor accuracy, NNMF gives rise to tuning properties that align more closely with those observed in primate MSTd than any of the accuracy-optimized CNNs. This work offers a step toward a deeper understanding of the computational properties and constraints that describe the optic flow tuning of primate area MSTd.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Optimizing accuracy and efficacy in data-driven materials discovery for the solar production of hydrogen

The production of hydrogen fuels, via water splitting, is of practical relevance for meeting global energy needs and mitigating the environmental consequences of fossil-fuel-based transportation. Water photoelectrolysis has been proposed as a viable approach for generating hydrogen, provided that stable and inexpensive photocatalysts with conversion efficiencies over 10% can be discovered, synthesized at scale, and successfully deployed. While a number of first-principles studies have focused on the data-driven discovery of photocatalysts, in the absence of systematic experimental validation, the success rate of these predictions may be limited. We address this problem by developing a screening procedure with co-validation between experiment and theory to expedite the synthesis, characterization, and testing of the computationally predicted, most desirable materials. Here, starting with 70 150 compounds in the Materials Project database, the proposed protocol yielded 71 candidate photocatalysts, 11 of which were synthesized as single-phase materials. Experiments confirmed hydrogen generation and favorable band alignment for 6 of the 11 compounds, with the most promising ones belonging to the families of alkali and alkaline-earth indates and orthoplumbates. This study shows the accuracy of a nonempirical, Hubbard-corrected density-functional theory method to predict band gaps and band offsets at a fraction of the computational cost of hybrid functionals, and outlines an effective strategy to identify photocatalysts for solar hydrogen generation.

08 HYDROGEN↗
DOE OSTI2015

Optimizing accuracy of determinations of CO₂ storage capacity and permanence, and designing more efficient storage operations: An example from the Rock Springs Uplift, Wyoming

At a potential injection site on the Rock Springs Uplift in southwest Wyoming, an investigation of confining layers was undertaken to develop and test methodology, identify key data requirements, assess previous injection scenarios relative to detailed confining layer properties, and integrate all findings in order to reduce the uncertainty of CO₂ storage permanence. The assurance of safe and permanent storage of CO₂ at a storage site involves a detailed evaluation of the confining layers. Four suites of field data were recognized as crucial for determining storage permanence relative to the confining layers; seismic, core and petrophysical data from a wellbore, formation fluid samples, and in-situ formation tests. Core and petrophysical data were used to create a vertical heterogenic property model that defined porosity, permeability, displacement pressure, geomechanical strengths, and diagenetic history. These analyses identified four primary confining layers and multiple redundant confining layers. In-situ formation tests were used to evaluate fracture gradients, regional stress fields, baseline microseismic data, step-rate injection tests, and formation perforation responses. Seismic attributes, correlated with the vertical heterogenic property models, were calculated and used to create a 3-D volume model over the entire site. The seismic data provided the vehicle to transform the vertical heterogenic property model into a horizontal heterogenic property model, which allowed for the evaluation of confining layers across the entire study site without risking additional wellbore perforations. Lastly, formation fluids were collected and analyzed for geochemical and isotopic compositions from stacked reservoir systems. These data further tested primary confining layers, by evaluating the evidence of mixing between target reservoirs (mixing would imply an existing breach of primary confining layers). All data were propagated into a dynamic, heterogenic geologic property model used to test various injection scenarios. These tests showed that the study site could retain 25MT of injected CO₂ over an injection lifespan of 50 years. Major findings indicate that active reservoir pressure management through reservoir fluid production (minimum of three production wells) greatly reduces the risk of breaching a confining layer. To address brine production, a well completion and engineering study was incorporated to reduce the risks of scaling and erosion during injection and production. These scenarios suggest that the dolostone within the Mississippian Madison Limestone is the site’s best injection/production target by two orders of magnitude, and that commercial well equipment would meet all performance requirements. This confirms that there are multiple confining layers in southwest Wyoming that are capable of retaining commercial volumes of CO₃, making Wyoming’s Paleozoic reservoirs ideal storage targets for low-risk injection and long-term storage. This study also indicates that column height retention calculations are reduced in a CO₂-brine system relative to a hydrocarbon-brine system, which is an observation that affects all potential CCS sites. Likewise, this study identified the impacts that downhole testing imparts on reservoir fluids, and the likelihood of introducing uncertainty in baseline site assumptions and later modeling.

25 ENERGY STORAGE↗
DOE OSTI2014

Chisel: reliability- and accuracy-aware optimization of approximate computational kernels

The accuracy of an approximate computation is the distance between the result that the computation produces and the corresponding fully accurate result. The reliability of the computation is the probability that it will produce an acceptably accurate result. Emerging approximate hardware platforms provide approximate operations that, in return for reduced energy consumption and/or increased performance, exhibit reduced reliability and/or accuracy. We present Chisel, a system for reliability- and accuracy-aware optimization of approximate computational kernels that run on approximate hardware platforms. Given a combined reliability and/or accuracy specification, Chisel automatically selects approximate kernel operations to synthesize an approximate computation that minimizes energy consumption while satisfying its reliability and accuracy specification. We evaluate Chisel on five applications from the image processing, scientific computing, and financial analysis domains. The experimental results show that our implemented optimization algorithm enables Chisel to optimize our set of benchmark kernels to obtain energy savings from 8.7% to 19.8% compared to the fully reliable kernel implementations while preserving important reliability guarantees.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Reduction of detection limit and quantification uncertainty due to interferent by neural classification with abstention

Many measurements in the physical sciences can be cast as counting experiments, where the number of occurrences of a physical phenomenon informs the prevalence of the phenomenon's source. Often, detection of the physical phenomenon (termed signal) is difficult to distinguish from naturally occurring phenomena (termed background). In this case, the discrimination of signal events from background can be performed using classifiers, and they may range from simple, threshold-based classifiers to sophisticated neural networks. These classifiers are often trained and validated to obtain optimal accuracy, however we show that the optimal accuracy classifier does not generally coincide with a classifier that provides the lowest detection limit, nor the lowest quantification uncertainty. Here, we present a derivation of the detection limit and quantification uncertainty in the classifier-based counting experiment case. We also present a novel abstention mechanism to minimize the detection limit or quantification uncertainty a posteriori. We illustrate the method on two data sets from the physical sciences, discriminating Ar-37 and Ar-39 radioactive decay from non-radioactive events in a gas proportional counter, and discriminating neutrons from gammas in an inorganic scintillator and report results therefrom.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2011

Optimizing the accuracy of a helical diode array dosimeter: A comprehensive calibration methodology coupled with a novel virtual inclinometer

Purpose: The goal of any dosimeter is to be as accurate as possible when measuring absolute dose to compare with calculated dose. This limits the uncertainties associated with the dosimeter itself and allows the task of dose QA to focus on detecting errors in the treatment planning (TPS) and/or delivery systems. This work introduces enhancements to the measurement accuracy of a 3D dosimeter comprised of a helical plane of diodes in a volumetric phantom. Methods: We describe the methods and derivations of new corrections that account for repetition rate dependence, intrinsic relative sensitivity per diode, field size dependence based on the dynamic field size determination, and positional correction. Required and described is an accurate ''virtual inclinometer'' algorithm. The system allows for calibrating the array directly against an ion chamber signal collected with high angular resolution. These enhancements are quantitatively validated using several strategies including ion chamber measurements taken using a ''blank'' plastic shell mimicking the actual phantom, and comparison to high resolution dose calculations for a variety of fields: static, simple arcs, and VMAT. A number of sophisticated treatment planning algorithms were benchmarked against ion chamber measurements for their ability to handle a large air cavity in the phantom. Results: Each calibration correction is quantified and presented vs its independent variable(s). The virtual inclinometer is validated by direct comparison to the gantry angle vs time data from machine log files. The effects of the calibration are quantified and improvements are seen in the dose agreement with the ion chamber reference measurements and with the TPS calculations. These improved agreements are a result of removing prior limitations and assumptions in the calibration methodology. Average gamma analysis passing rates for VMAT plans based on the AAPM TG-119 report are 98.4 and 93.3% for the 3%/3 mm and 2%/2 mm dose-error/distance to agreement threshold criteria, respectively, with the global dose-error normalization. With the local dose-error normalization, the average passing rates are reduced to 94.6 and 85.7% for the 3%/3 mm and 2%/2 mm criteria, respectively. Some algorithms in the convolution/superposition family are not sufficiently accurate in predicting the exit dose in the presence of a 15 cm diameter air cavity. Conclusions: Introduction of the improved calibration methodology, enabled by a robust virtual inclinometer algorithm, improves the accuracy of the dosimeter's absolute dose measurements. With our treatment planning and delivery chain, gamma analysis passing rates for the VMAT plans based on the AAPM TG-119 report are expected to be above 91% and average at about 95% level for {gamma}(3%/3 mm) with the local dose-error normalization. This stringent comparison methodology is more indicative of the true VMAT system commissioning accuracy compared to the often quoted dose-error normalization to a single high value.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

Optimizing the accuracy of viscoelastic characterization with AFM force–distance experiments in the time and frequency domains

Atomic Force Microscopy (AFM) force-distance (FD) experiments have emerged as an attractive alternative to traditional micro-rheology measurement techniques owing to their versatility of use in materials of a wide range of mechanical properties. Here, we show that the range of time dependent behaviour which can reliably be resolved from the typical method of FD inversion (fitting constitutive FD relations to FD data) is inherently restricted by the experimental parameters: sampling frequency, experiment length, and strain rate. Specifically, we demonstrate that violating these restrictions can result in errors in the values of the parameters of the complex modulus. In the case of complex materials, such as cells, whose behaviour is not specifically understood a priori, the physical sensibility of these parameters cannot be assessed and may lead to falsely attributing a physical phenomenon to an artifact of the violation of these restrictions. We use arguments from information theory to understand the nature of these inconsistencies as well as devise limits on the range of mechanical parameters which can be reliably obtained from FD experiments. The results further demonstrate that the nature of these restrictions depends on the domain (time or frequency) used in the inversion process, with the time domain being far more restrictive than the frequency domain. Lastly, we demonstrate how to use these restrictions to better design FD experiments to target specific timescales of a material's behaviour through our analysis of a polydimethylsiloxane (PDMS) polymer sample.

information theory↗
DOE OSTI2014

Reliable clock estimation using linear weighted fusion based on pairwise broadcast synchronization

Clock synchronization in wireless sensor networks (WSNs) has been studied extensively in recent years and many protocols are put forward based on the point of statistical signal processing, which is an effective way to optimize accuracy. However, the accuracy derived from the statistical data can be improved mainly by sufficient packets exchange, which will consume the limited power resources greatly. In this paper, a reliable clock estimation using linear weighted fusion based on pairwise broadcast synchronization is proposed to optimize sync accuracy without expending additional sync packets. As a contribution, a linear weighted fusion scheme for multiple clock deviations is constructed with the collaborative sensing of clock timestamp. And the fusion weight is defined by the covariance of sync errors for different clock deviations. Extensive simulation results show that the proposed approach can achieve better performance in terms of sync overhead and sync accuracy.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2020

Variable-fidelity multipoint aerodynamic shape optimization with output-based adapted meshes

This work presents a method to control the discretization error in multipoint aerodynamic shape optimization using output-based adapted meshes. The meshes are adapted via adjoint-based error estimates, taking into account both the objective and constraint output errors. A multi-fidelity optimization framework is then developed by taking advantage of the variable fidelity offered by adaptive meshes. The objective functional and its sensitivity at each design point (operating condition) are first evaluated on the same initial coarse mesh, which is then subsequently adapted for each design point individually as the shape optimization proceeds. The effort to set up the optimization is minimal since the initial mesh can be fairly coarse and easy to generate. As the shape approaches the optimal design, the mesh at each design point becomes finer, in regions necessary for that particular operating condition. The multi-fidelity framework is tightly coupled with the objective error estimation to ensure the optimization accuracy at each fidelity. Computational savings arise from a reduction of the mesh size when the design is far from optimal and avoiding an exhaustive search on low-fidelity meshes. The proposed method is demonstrated on multipoint drag minimization problems of a transonic airfoil with lift and area constraints. Improved accuracy and efficiency are shown compared to traditional fixed-fidelity optimization with a fixed computational mesh.

42 ENGINEERING↗
DOE OSTI2026

Large Language Model for Validation, Optical Calibration, and Learning (VOCAL) Distributed Temperature Sensing Interface

Distributed temperature sensing (DTS) using fiber optic sensors (FOS) offers a promising method for temperature measurements in advanced reactors, such as sodium fast reactors and molten salt cooled reactors. To support the calibration and validation of DTS measurements, Argonne National Laboratory developed the Validation, Optical Calibration, and Learning (VOCAL) software package. This report describes the integration of a local large language model (LLM) with a retrieval-augmented generation (RAG) system into the VOCAL interface to serve as an interactive user assistant. The LLM framework enhances the VOCAL platform’s accessibility to users by explaining interface components, clarifying inputs and outputs, and answering user queries dynamically in real-time. The accuracy of the LLM assistant performance was evaluated with 20 queries regarding the interface and its parameters using experimental data from the Thermal Hydraulic Experimental Test Article (THETA) facility. Results demonstrate that the LLM achieved a 95% accuracy rate, with a BERTScore of 0.8816 and SBERT value of 0.7417. Furthermore, validation of the RAG system within the LLM framework showed optimal accuracy with k-values between 1 and 2 using the k-refinement convergence test. The prompt perturbation analysis demonstrated good initial consistency for the RAG system, exhibiting the highest accuracy under punctuation variations and the greatest sensitivity under query reordering. Notably, the model’s errors were limited to data retrieval failures rather than factual hallucinations, reinforcing its baseline reliability. The integration of LLM provides a highly accurate, userfriendly enhancement to the VOCAL platform without disrupting its core computational capabilities for FOS calibration and validation.

Hong, Evan↗
DOE OSTI2018

Furthering Precision in Sentinel Node Navigational Surgery for Oral Cancer: a Novel Triple Targeting System

To describe an innovative sentinel lymph node (SLN) guidance approach using a radionuclide tracer, 3D augmented reality-guided imaging, and near infrared (NIR) fluorescence over-lay imaging with hand-held probes to optimize accuracy, efficiency, and precise navigation for sentinel node (SN) localization in head and neck cancer. In a cT1N0M0 squamous cell carcinoma of the tongue, pre-operative radionuclide lymphoscintigraphy was performed with a sentinel node-specific radiolabeled tracer. Intraoperatively, a 3D hand-held augmented reality (AR) scanning SPECT probe assessed concordance of the SN with pre-operative SPECT-CT images. The real-time optical video was linked to the SPECT-CT images for added precision. Final guidance to the SN was performed using ICG fluorescence imaging. Dynamic and SPECT-CT showed bilateral lymphatic drainage from the tumor. The 3D hand-held AR SPECT probe SN localization was concordant with pre-operative imaging. The optical video successfully demonstrated the lymphatic drainage in real-time through a unique overlay fluorescence image. The ICG localized to the same nodes identified by both the SPECT-CT and hand-held SPECT images. The use of dual radiation and fluorescence tracers improved SN detection, especially for SN close to the injection site. The hand-held probes allowed the surgeon to dissect continuously, without needing to change tools. The combination of augmented reality, nuclear medicine, and over-lay fluorescence imaging allowed greater accuracy for matching the preoperative imaging with intraoperative identification and precisely guiding the dissection. This method uniquely permitted the surgeon to efficiently dissect the SN with accurate visualization and optimal precision.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2024

Nonlinear elasticity with the Shifted Boundary Method

Here, we propose a new unfitted/immersed computational framework for nonlinear solid mechanics, which bypasses the complexities associated with the generation of CAD representations and subsequent body-fitted meshing. This approach allows to speed up the cycle of design and analysis in complex geometry and requires relatively simple computer graphics representations of the surface geometries to be simulated, such as the Standard Tessellation Language (STL format). Complex data structures and integration on cut elements are avoided by means of an approximate boundary representation and a modification (shifting) of the boundary conditions to maintain optimal accuracy. An extensive set of computational experiments in two and three dimensions is included.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Matrix-free subcell residual distribution for Bernstein finite elements: Monolithic limiting

This paper is focused on the aspects of limiting in residual distribution (RD) schemes for high-order finite element approximations to advection problems. Both continuous and discontinuous Galerkin methods are considered in this work. Discrete maximum principles are enforced using algebraic manipulations of element contributions to the global nonlinear system. The required modifications can be carried out without calculating the element matrices and assembling their global counterparts. The components of element vectors associated with the standard Galerkin discretization are manipulated directly using localized subcell weights to achieve optimal accuracy. Low-order nonlinear RD schemes of this kind were originally developed to calculate local extremum diminishing predictors for flux-corrected transport (FCT) algorithms. In the present paper, we incorporate limiters directly into the residual distribution procedure, which makes it applicable to stationary problems and leads to well-posed nonlinear discrete problems. To circumvent the second-order accuracy barrier, the correction factors of monolithic limiting approaches and FCT schemes are adjusted using smoothness sensors based on second derivatives. Finally, the convergence behavior of presented methods is illustrated by numerical studies for two-dimensional test problems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

EHR-BERT: A BERT-based model for effective anomaly detection in electronic health records

Objective: Physicians and clinicians rely on data contained in electronic health records (EHRs), as recorded by health information technology (HIT), to make informed decisions about their patients. The reliability of HIT systems in this regard is critical to patient safety. Consequently, better tools are needed to monitor the performance of HIT systems for potential hazards that could compromise the collected EHRs, which in turn could affect patient safety. In this paper, we propose a new framework for detecting anomalies in EHRs using sequence of clinical events. This new framework, EHR-Bidirectional Encoder Representations from Transformers (BERT), is motivated by the gaps in the existing deep-learning related methods, including high false negatives, sub-optimal accuracy, higher computational cost, and the risk of information loss. EHR-BERT is an innovative framework rooted in the BERT architecture, meticulously tailored to navigate the hurdles in the contemporary BERT method; thus, enhancing anomaly detection in EHRs for healthcare applications.Methods: The EHR-BERT framework was designed using the Sequential Masked Token Prediction (SMTP) method. This approach treats EHRs as natural language sentences and iteratively masks input tokens during both training and prediction stages. This method facilitates the learning of EHR sequence patterns in both directions for each event and identifies anomalies based on deviations from the normal execution models trained on EHR sequences.Results: Extensive experiments on large EHR datasets across various medical domains demonstrate that EHR-BERT markedly improves upon existing models. It significantly reduces the number of false positives and enhances the detection rate, thus bolstering the reliability of anomaly detection in electronic health records. This improvement is attributed to the model’s ability to minimize information loss and maximize data utilization effectively.Conclusion: EHR-BERT showcases immense potential in decreasing medical errors related to anomalous clinical events, positioning itself as an indispensable asset for enhancing patient safety and the overall standard of healthcare services. The framework effectively overcomes the drawbacks of earlier models, making it a promising solution for healthcare professionals to ensure the reliability and quality of health data.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗