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 1,459 records · Page 81

Hybrid Modeling for Complex Systems Health Management

The research work presents application of hybrid physics-informed machine learning to a representative electric powertrain for unmanned aerial vehicles. The model is composed of physics-derived and empirical equations, integrated with connected networks that are strategically placed within the model to substitute equations that are subject to large uncertainty. Polynomial fit driven by heuristics or empirical observations can be substituted by more flexible networks that can minimize the error between model predictions and observations without being restricted to a predefined functional form. This modeling strategy allows training of networks deep inside the model and unknown parameters in a single learning stage. The powertrain model consists of Li-ion batteries, electronic speed controller with pulse-width modulation, and brush-less DC motor with connected propeller. Results obtained from combination of laboratory and simulation tests are discussed in this work.

PINNS↗

Hybrid Modeling of Unmanned Aerial Vehicle Electric Powertrain for Fault Detection and Diagnostics

This paper shows the application of hybrid physics-informed machine learning to a representative electric powertrain for unmanned aerial vehicles. The model is composed of physics-derived principles and empirical equations, as well as fully connected networks that are strategically placed within the model to substitute equations that are subject to large uncertainty. Polynomial fit driven by heuristics or empirical observations can be substituted by more flexible networks that can minimize the error between model predictions and observations without being restricted to a predefined functional form. This modeling strategy allows training of networks deep inside the model and unknown parameters in a single learning stage. It has already been applied to Li-ion batteries in the past, and in this work, we extend the applications to other components of an electric powertrain, namely electronic speed controller with pulse-width modulation, and brushless DC motor with connected propeller. Training and testing of the model is carried out using experimental data from Li-ion battery discharge and powertrain testing in a laboratory environment.

Physics-Informed Machine Learning↗

Physics Informed Neural Nets for Systems Health Management

To facilitate and solve the prediction problem, awareness of the current health state of the system is key, since it is necessary to perform condition-based predictions. To accurately predict the future state of any system, it is required to possess knowledge of its current health state and future operational conditions. Development in data-driven algorithms in regression of complex nonlinear functions and classification tasks have generated a growing interest in artificial intelligence for industrial applications. Complex multi-physics models as well as digital twins, once purely built on physics and corresponding simplified lumped parameter iterations, can now benefit from machine learning algorithms to mitigate the lack of understanding of some complex behavior. The research work presents application of physics-informed neural nets application to a representative electric powertrain for unmanned aerial vehicles. The model is composed of physics-derived and empirical equations, integrated with connected networks that are strategically placed within the model to substitute equations that are subject to large uncertainty. Polynomial fit driven by heuristics or empirical observations can be substituted by more flexible networks that can minimize the error between model predictions and observations without being restricted to a predefined functional form. This modeling strategy allows training of networks deep inside the model and unknown parameters in a single learning stage.

Physics Informed↗

Model Based Approaches for Fault Detection, Prognostics, Decision Making in Complex Systems

The presentation discusses application of model based approaches to complex systems. The model is composed of physics-derived and empirical equations, integrated with connected networks that are strategically placed within the model to substitute equations that are subject to large uncertainty. Polynomial fit driven by heuristics or empirical observations can be substituted by more flexible networks that can minimize the error between model predictions and observations without being restricted to a predefined functional form. This modeling strategy allows training of networks deep inside the model and unknown parameters in a single learning stage.

Physics Informed↗

Software Model Of Software-Development Process

Collection of computer programs constitutes software tool for simulation of medium- to large-scale software-development projects. Necessary to include easily identifiable and more-readily quantifiable characteristics like costs, times, and numbers of errors. Mathematical model incorporating these and other factors of dynamics of software-development process implemented in the Software Life Cycle Simulator (SLICS) computer program. Simulates dynamics of software-development process. In combination with input and output expert software systems and knowledge-based management software system, develops information for use in managing large software-development project. Intended to aid managers in planning, managing, and controlling software-development processes by reducing uncertainties in budgets, required personnel, and schedules.

Lin, Chi Y.↗

Domain-aware Control-oriented Neural Models for Autonomous Underwater Vehicles

Conventional physics-based modeling is a time-consuming bottleneck in control design for complex nonlinear systems like autonomous underwater vehicles (AUVs). In contrast, purely data-driven models, require a large number of observations and lack operational guarantees for safety-critical systems. Data-driven models leveraging available partially characterized dynamics have potential to provide reliable systems models in a typical data-limited scenario for high value complex systems, thereby avoiding months of expensive expert modeling time. In this work we explore this middle-ground between expert-modeled and pure data-driven modeling. We present control-oriented parametric models with varying levels of domain-awareness that exploit known system structure and prior physics knowledge to create constrained deep neural dynamical system models. We employ universal differential equations to construct data-driven blackbox and graybox representations of the AUV dynamics. In addition, we explore a hybrid formulation that explicitly models the residual error related to imperfect graybox models. We compare the prediction performance of the learned models for different distributions of initial conditions and control inputs to assess their suitability for control.

Shaw Cortez, Wenceslao E.↗

Reduced models for ETG transport in the tokamak pedestal

This paper reports on the development of reduced models for electron temperature gradient (ETG) driven transport in the pedestal. Model development is enabled by a set of 61 nonlinear gyrokinetic simulations with input parameters taken from pedestals in a broad range of experimental scenarios. The simulation data have been consolidated in a new database for gyrokinetic simulation data, the multiscale gyrokinetic database (MGKDB), facilitating the analysis. The modeling approach may be considered a generalization of the standard quasilinear mixing length procedure. The parameter η, the ratio of the density to temperature gradient scale length, emerges as the key parameter for formulating an effective saturation rule. With a single order-unity fitting coefficient, the model achieves an error of 15%. A similar model for ETG particle flux is also described. We also present simple algebraic expressions for the transport informed by an algorithm for symbolic regression.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Prediction of hydration energies of adsorbates at Pt(111) and liquid water interfaces using machine learning

Aqueous phase heterogeneous catalysis is important to various industrial processes, including biomass conversion, Fischer–Tropsch synthesis, and electrocatalysis. Accurate calculation of solvation thermodynamic properties is essential for modeling the performance of catalysts for these processes. Explicit solvation methods employing multiscale modeling, e.g., involving density functional theory and molecular dynamics have emerged for this purpose. Although accurate, these methods are computationally intensive. This study introduces machine learning (ML) models to predict solvation thermodynamics for adsorbates on a Pt(111) surface, aiming to enhance computational efficiency without compromising accuracy. In particular, ML models are developed using a combination of molecular descriptors and fingerprints and trained on previously published water–adsorbate interaction energies, energies of solvation, and free energies of solvation of adsorbates bound to Pt(111). These models achieve root mean square error values of 0.09 eV for interaction energies, 0.04 eV for energies of solvation, and 0.06 eV for free energies of solvation, demonstrating accuracy within the standard error of multiscale modeling. Feature importance analysis reveals that hydrogen bonding, van der Waals interactions, and solvent density, together with the properties of the adsorbate, are critical factors influencing solvation thermodynamics. Furthermore, these findings suggest that ML models can provide rapid and reliable predictions of solvation properties. This approach not only reduces computational costs but also offers insights into the solvation characteristics of adsorbates at Pt(111)–water interfaces.

Adsorption↗

Bayesian model averaging for analysis of lattice field theory results

Statistical modeling is a key component in the extraction of physical results from lattice field theory calculations. Although the general models used are often strongly motivated by physics, many model variations can frequently be considered for the same lattice data. Model averaging, which amounts to a probability-weighted average over all model variations, can incorporate systematic errors associated with model choice without being overly conservative. We discuss the framework of model averaging from the perspective of Bayesian statistics, and give useful formulae and approximations for the particular case of least-squares fitting, commonly used in modeling lattice results. In addition, we frame the common problem of data subset selection (e.g. choice of minimum and maximum time separation for fitting a two-point correlation function) as a model selection problem and study model averaging as a straightforward alternative to manual selection of fit ranges. Numerical examples involving both mock and real lattice data are given.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Improved Parametric Models for Explosion Pressure Signals Derived From Large Datasets

Accurate recording and characterization of explosion-induced pressure signals are key components of the forensic analysis of explosion events in the atmosphere. Parametric overpressure models based on several key waveform features (peak overpressure, positive pulse duration, and impulse) are widely used to estimate explosion energy in terms of trinitrotoluene equivalent yield. However, those models are often developed by a limited dataset, including only a few events or recordings at relatively short propagation distances. Here, we develop empirical waveform-parameter models based on a regression analysis of a large set of data curated from four chemical explosion experiments including 16 detonations. We measured peak overpressure and impulse for positive and negative phases from 1000 pressure signals recorded at local ranges (<20 km) with scaled distance up to 8000 m/kg 1/3 . Additionally, the measured waveform parameters showed large variation with respect to observing distances indicating the effects of atmospheric propagation. In this study, a second-order polynomial model was used in a least-squares regression to account for those propagation effects and to improve data fitting. In addition to model parameters for waveform features, we also determined range-dependent model uncertainties based on data variance. The model uncertainty represents the prediction error of our models and can be critical to evaluating the uncertainty of yield estimate.

58 GEOSCIENCES↗

High Accuracy Transistor Compact Model Calibrations

Typically, transistors are modeled by the application of calibrated nominal and range models. These models consists of differing parameter values that describe the location and the upper and lower limits of a distribution of some transistor characteristic such as current capacity. Correspondingly, when using this approach, high degrees of accuracy of the transistor models are not expected since the set of models is a surrogate for a statistical description of the devices. The use of these types of models describes expected performances considering the extremes of process or transistor deviations. In contrast, circuits that have very stringent accuracy requirements require modeling techniques with higher accuracy. Since these accurate models have low error in transistor descriptions, these models can be used to describe part to part variations as well as an accurate description of a single circuit instance. Thus, models that meet these stipulations also enable the calculation of quantification of margins with respect to a functional threshold and uncertainties in these margins. Given this need, new model high accuracy calibration techniques for bipolar junction transistors have been developed and are described in this report.

42 ENGINEERING↗

Computer Aided Modeling and Post Processing with NASTRAN Analysis

Computer aided engineering systems are invaluable tools in performing NASTRAN finite element analysis. These techniques are implemented in both the pre-processing and post-processing phases of the NASTRAN analysis. The finite element model development, or pre-processing phase, was automated with a computer aided modeling program called Supertabl, and the review and interpretation of the results of the NASTRAN analysis, or post-processing phase, was automated with a computer aided plotting program called Output Display. An intermediate program, Nasplot, which was developed in-house, has also helped to cut down on the model checkout time and reduce errors in the model. An interface has been established between the finite element computer aided engineering system and the Learjet computer aided design system whereby data can be transferred back and forth between the two. These systems have significantly improved productivity and the ability to perform NASTRAN analysis in response to product development requests.

Boroughs, R. R.↗

Error Generation in CATS-Based Agents

This research presents a methodology for generating errors from a model of nominally preferred correct operator activities, given a particular operational context, and maintaining an explicit link to the erroneous contextual information to support analyses. It uses the Crew Activity Tracking System (CATS) model as the basis for error generation. This report describes how the process works, and how it may be useful for supporting agent-based system safety analyses. The report presents results obtained by applying the error-generation process and discusses implementation issues. The research is supported by the System-Wide Accident Prevention Element of the NASA Aviation Safety Program.

Callantine, Todd↗

Application of a Multiscale Model of Tantalum Deformation at Megabar Pressures

A new multiscale simulation tool has been developed to model the strength of tantalum under high-pressure dynamic compression. This new model combines simulations at multiple length scales to explain macroscopic properties of materials. Previously known continuum models of material response under load have built upon a mixture of theoretical physics and experimental phenomenology. Experimental data, typically measured at static pressures, are used as a means of calibration to construct models that parameterize the material properties; e.g., yield stress, work hardening, strain-rate dependence, etc. The pressure dependence for most models enters through the shear modulus, which is used to scale the flow stress. When these models are applied to data taken far outside the calibrated regions of phase space (e.g., strain rate or pressure) they often diverge in their predicted behavior of material deformation. The new multiscale model, developed at Lawrence Livermore National Laboratory, starts with interatomic quantum mechanical potential and is based on the motion and multiplication of dislocations. The basis for the macroscale model is plastic deformation by phonon drag and thermally activated dislocation motion and strain hardening resulting from elastic interactions among dislocations. The dislocation density, {rho}, and dislocation velocity, {nu}, are connected to the plastic strain rate {var_epsilon}{sup p}, via Orowan's equation: {var_epsilon}{sup p} = {rho}b{nu}/M, where b is the Burger's vector, the shear magnitude associated with a dislocation, and M is the Taylor factor, which accounts for geometric effects in how slip systems accommodate the deformation. The evolution of the dislocation density and velocity is carried out in the continuum model by parameterized fits to smaller scale simulations, each informed by calculations on smaller length scales down to atomistic dimensions. We apply this new model for tantalum to two sets of experiments and compare the results with more traditional models. The experiments are based on the Barnes's technique in which a low density material loads against a metal surface containing a pre-imposed rippled pattern. The loaded sample is Rayleigh-Taylor unstable and the rippled amplitudes grow with time. The rate of growth differs depending on the material strength, with stronger materials growing less, even to the point of saturation. One set of experiments was conducted at the pRad facility at LANSCE at Los Alamos National Laboratory in 2007 using high-explosive (HE) driven tantalum samples. The other set of experiments was done at the Omega laser at the Laboratory for Laser Energetics at the University of Rochester, which used high-powered lasers to create plasmas to dynamically compress a rippled tantalum sample. The two techniques provide data at different pressures and strain rates: The HE technique drives the samples at around 2 x 10{sup 5} s{sup -1} strain rate and pressures near 500 kbar, while the laser technique hits strain rates around 2 x 10{sup 7} s{sup -1} and pressures close to 1.4 Mbar. The most recent laser experiments were conducted in February 2010 and they present a sample of the data in Figure 1, which shows a face-on radiograph at a time of 65 ns after the laser was turned on. From this radiograph, they measure the growth factor which is defined to be the change in amplitude of the ripples relative to their initial amplitude. Figure 2 shows the resulting growth factors along with various model fits. The error bars are typically 20-25%. Only the multiscale model predictions match the experimental measurements. The growth factors via the HE technique are determined from multiple side-on proton radiography images and thus provide a full growth curve per single experiment. A sample growth curve is shown in Figure 3, also with various model fits and error bars estimated at 25%. It should be noted that by 7.5 {micro}s the growth in this sample has exceeded the initial target thickness indicating that localizations not captured in the overall simulation have probably become dominant, i.e., the target is likely breaking up. Application of the multiscale dislocation dynamics model as implemented in the Ares hydrodynamics code shows excellent agreement with both the pRad and Omega data. They also compare the Steinberg-Lund (SL), Preston-Tonks-Wallace (PTW), and Stainberg-Guinan (SG) models with the data. The PTW and SG models provide good fits to the pRad data but over-predict the growth (underestimate the strength) on the laser platform. The SL model under-predicts the pRad data and over-predicts the Omega data. The excellent agreement of the multiscale model with the data over two orders of magnitude in strain rate and more than a factor of two in pressure lends credibility to the model. They continue to stress the model by conducting experiments at 5 Mbars and beyond at the National Ignition Facility at LLNL in the near future.

36 MATERIALS SCIENCE↗

The Diurnal Cycle of Clouds and Precipitation at the ARM SGP Site: An Atmospheric State-Based Analysis and Error Decomposition of a Multiscale Modeling Framework Simulation

Long–term reflectivity data collected by a millimeter cloud radar at the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site are used to examine the diurnal cycle of clouds and precipitation and are compared with the diurnal cycle simulated by a Multiscale Modeling Framework (MMF) climate model. The study uses a set of atmospheric states that were created specifically for the SGP and for the purpose of investigating under what synoptic conditions models compare well with observations on a statistical basis (rather than using case studies or seasonal or longer time scale averaging). Differences in the annual mean diurnal cycle between observations and the MMF are decomposed into differences due to the relative frequency of states, the daily mean vertical profile of hydrometeor occurrence, and the (normalized) diurnal variation of hydrometeors in each state. Here the hydrometeors are classified as cloud or precipitation based solely on the reflectivity observed by a millimeter radar or generated by a radar simulator. The results show that the MMF does not capture the diurnal variation of low clouds well in any of the states but does a reasonable job capturing the diurnal variations of high clouds and precipitation in some states. In particular, the diurnal variations in states that occur during summer are reasonably captured by the MMF, while the diurnal variations in states that occur during the transition seasons (spring and fall) are not well captured. Overall, the errors in the annual composite are due primarily to errors in the daily mean of hydrometeor occurrence (rather than diurnal variations), but errors in the state frequency (that is, the distribution of weather states in the model) also play a significant role.

54 ENVIRONMENTAL SCIENCES↗

Model-based analysis of the impact of diffuse radiation on CO 2 exchange in a temperate deciduous forest

Clouds and aerosols increase the fraction of global solar irradiance that is diffuse light. This phenomenon is known to increase the photosynthetic light use efficiency (LUE) of closed-canopy vegetation by redistributing photosynthetic photon flux density (400–700 nm) from saturated, sunlit leaves at the top of the canopy, to shaded leaves deeper in the canopy. We combined a process-based carbon cycle model with 10 years of eddy covariance carbon flux measurements and other ancillary data sets to assess 1) how this LUE enhancement influences interannual variation in carbon uptake, and 2) how errors in modeling diffuse fraction affect predictions of carbon uptake. Modeled annual gross primary productivity (GPP) increased by ≈0.94% when observed levels of diffuse fraction were increased by 0.01 (holding total irradiance constant). The sensitivity of GPP to increases in diffuse fraction was highest when the diffuse fraction was low to begin with, and lowest when the diffuse fraction was already high. Diffuse fraction also explained significantly more of the interannual variability of modeled net ecosystem exchange (NEE), than did total irradiance. Two tested radiation partitioning models yielded over- and underestimates of diffuse fraction at our site, which propagated to over- and underestimates of annual NEE, respectively. Our findings highlight the importance of incorporating LUE enhancement under diffuse light into models of global primary production, and improving models of diffuse fraction.

54 ENVIRONMENTAL SCIENCES↗

Simulations of tokamak edge plasma turbulent fluctuations based on a minimal 3D model

In this study, a new simulation model for tokamak boundary plasma, SOLT3D, is implemented in the BOUT++ framework. The simulation model includes a set of dynamic equations describing collisional boundary plasma and neutral gas in the tokamak scrape-off layer and divertor region. The model is verified against standard linear plasma instabilities and available nonlinear results. For L-mode like conditions, SOLT3D reproduces characteristics of boundary plasma turbulent fluctuations that are similar to published experimental data, in terms of the amplitude and spatial dependence of the fluctuations. It also reproduces realistic plasma fluxes on material surfaces and expected Bohm-like effective radial transport. Plasma fluctuations inferred from the simulations imply inevitably a significant level of intrinsic 'noise' for axisymmetric tokamak plasma transport modeling, introducing errors on the order of unity. In particular, the toroidally averaged atomic rates below 5–10 eV are strongly modified by turbulent plasma fluctuations, which should significantly affect the standard axisymmetric modeling of the tokamak edge plasma and divertor.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Man-Machine Integration Design and Analysis System (MIDAS) v5: Augmentations, Motivations, and Directions for Aeronautics Applications

As automation and advanced technologies are introduced into transport systems ranging from the Next Generation Air Transportation System termed NextGen, to the advanced surface transportation systems as exemplified by the Intelligent Transportations Systems, to future systems designed for space exploration, there is an increased need to validly predict how the future systems will be vulnerable to error given the demands imposed by the assistive technologies. One formalized approach to study the impact of assistive technologies on the human operator in a safe and non-obtrusive manner is through the use of human performance models (HPMs). HPMs play an integral role when complex human-system designs are proposed, developed, and tested. One HPM tool termed the Man-machine Integration Design and Analysis System (MIDAS) is a NASA Ames Research Center HPM software tool that has been applied to predict human-system performance in various domains since 1986. MIDAS is a dynamic, integrated HPM and simulation environment that facilitates the design, visualization, and computational evaluation of complex man-machine system concepts in simulated operational environments. The paper will discuss a range of aviation specific applications including an approach used to model human error for NASA s Aviation Safety Program, and what-if analyses to evaluate flight deck technologies for NextGen operations. This chapter will culminate by raising two challenges for the field of predictive HPMs for complex human-system designs that evaluate assistive technologies: that of (1) model transparency and (2) model validation.

Gore, Brian F.↗