Search NASA⌕ Search

SEARCH · Search NASA

Results for “Predictive modeling”

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 1,261 records · Page 70

Fouling modeling and prediction approach for heat exchangers using deep learning

In this article, we develop a generalized and scalable statistical model for accurate prediction of fouling resistance using commonly measured parameters of industrial heat exchangers. This prediction model is based on deep learning where a scalable algorithmic architecture learns non-linear functional relationships between a set of target and predictor variables from large number of training samples. Here, the efficacy of this modeling approach is demonstrated for predicting fouling in an analytically modeled cross-flow heat exchanger, designed for waste heat recovery from flue-gas using room temperature water. The performance results of the trained models demonstrate that the mean absolute prediction errors are under 10 –4 KW –1 for flue-gas side, water side and overall fouling resistances. The coefficients of determination (R 2 ), which characterize the goodness of fit between the predictions and observed data, are over 99%. Even under varying levels of measurement noise in the inputs, we demonstrate that predictions over an ensemble of multiple neural networks achieves better accuracy and robustness to noise. We find that the proposed deep-learning fouling prediction framework learns to follow heat exchanger flow and heat transfer physics, which we confirm using locally interpretable model agnostic explanations around randomly selected operating points. Overall, we provide a robust algorithmic framework for fouling prediction that can be generalized and scaled to various types of industrial heat exchangers.

42 ENGINEERING↗

Developing an Automated Uncertainty Quantification Tool to Improve Watershed-Scale Predictions of Water and Nutrient Cycling

Managing the flow of water, nutrients, and contaminants in watersheds is vital to addressing pressing issues related to water scarcity, access to clean drinking water, energy production, resilience to natural and anthropogenic perturbations, and ecological restoration. Decisions about the management of watersheds critically depend on the accuracy with which the flow of water and chemicals through the watershed can be predicted by computer models. Prediction uncertainty can be reduced by matching the model to data, which are collected in the field at great expense. The contribution of watershed characterization data to reducing uncertainty of relevant model predictions can be evaluated in a so-called data-worth analysis, which provides transparent, quantitative metrics about a data set’s value for the support of relevant watershed management objectives. To achieve this goal, we developed a software package that implements the data-worth analysis approach for use with state-of-the-art watershed models. The purpose of the proposed data-worth analysis is to help decision-makers allocate resources for watershed characterization such that the uncertainty in model predictions can be significantly reduced, which leads to better, more effective management decisions. At the same time, watershed characterization costs can be reduced. The specific technical objectives of this SBIR/STTR Phase II project were to develop a framework and associated software toolsets that implement the uncertainty quantification and data-worth analysis approach for use with state-of-the-art watershed models. This goal was achieved by (A) developing a user-friendly, robust software package that is accessible to a wide audience, including watershed managers, policy-makers, and public stakeholders; (B) by demonstrating application of the prototype on several use cases that are representative of complex watershed management challenges spanning a range of scales and that consider different open-source, DOE-based codes and other modeling platforms; and (C) by gathering information about the needs and requirements from potential users to help guide future developments, ensuring that the final product will be commercially viable. The developed software consists of a graphical user interface that guides the user through a sequence of analysis steps, supported by toolsets that leverage state-of-the-art computational simulation-optimization capabilities. A prototype of the software runs on multiple platforms (PC, Mac, multi-processor Linux environment), is linked to diverse watershed simulators (e.g., ECOSYS, TOUGH2, TOUGHREACT, Amanzi-ATS), performs multiple analysis tasks (predictive simulations, sensitivity analysis, uncertainty analysis, automatic parameter estimation, and data-worth analysis, multicomponent geothermometry), and is readily extensible to include external simulators and analysis tools. The software is being commercialized and will be continually updated to address user needs.

58 GEOSCIENCES↗

Ocean-Science Mission Needs: Real-Time AUV Data for Command, Control, and Model Inputs

Predictive models for tides, hydrodynamics, and bio-optical properties affecting the visibility and buoyancy of coastal waters are needed to evaluate the safety of personnel and equipment engaged in maritime operations under potentially hazardous conditions. Predicted currents can be markedly different for two-layer systems affected by terrestrial runoff than for well-mixed conditions because the layering decouples the surface and bottom Ekman layers and rectifies the current response to oscillatory upwelling-and downwelling-favorable winds. Standard ocean models (e.g. Princeton Ocean Model) require initial-and boundary data on the physical and optical properties of the multilayered water column to provide accurate simulations of heat budgets and circulation. Two observational systems are designed to measure vertically structured conditions on the West Florida Shelf (WFS): a tethered buoy network and an autonomous underwater vehicle (AUV) observational system. The AUV system is described with a focus on the observational systems that challenge or limit the communications command and control network for various types of measurement programs. These include vertical oscillatory missions on shelf transects to observe the optical and hydrographic properties of the water column, and bottom-following missions for measuring the bottom albedo. Models of light propagation, absorption, and conversion to heat as well as determination of the buoyancy terms for physical models require these measurements. High data rates associated with video bottom imagery are the most challenging for the real-time, command and control communications system, but they are met through a combination of loss-less and lossy data-compression methods, depending upon the data-rate of the radio links.

Carder, Kendall L.↗

Tripled yield in direct-drive laser fusion through statistical modelling

Accurate predictive models are essential to the design of high-performance, laser-fusion experiments. Here, we report the development of a promising statistical approach that has been used to quantitatively predict and design several cryogenic deuterium-tritium (DT) implosions on the OMEGA Laser System, leading to the tripling of the fusion yield to its highest value for direct-drive laser fusion of DT-layered targets. When scaled to National Ignition Facility laser energies of 1.9 MJ, these targets are predicted to produce a fusion energy output of ~500 kJ—several times larger than current fusion yields achieved at that facility. Future applications of this approach will guide the exploration of the vast parameter space for the conditions required for thermonuclear ignition and expose gaps in our understanding of laser-fusion physics.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

The Cell Utilized Partitioning Model as a Predictive Tool for Optimizing Counter-Current Chromatography Processes

Counter-current chromatography (CCC) is capable of unique elution modes that isolate analytes using the movement of the stationary phase in addition to moving the mobile phase. These modes include elution-extrusion CCC (EECCC) and dual-mode CCC (DM CCC) that are not possible in traditional solid-liquid chromatography systems. Although EECCC and DM CCC are widely used to recover highly retained components, to our knowledge, optimizing the elution process in these modes with predictive models has not been reported. To address this gap, we developed a predictive model for CCC dubbed the Cell Utilized Partitioning (CUP) model. The CUP model accurately predicts the effluents of multicomponent separations in EECCC and DM CCC modes when compared to experimental data. Furthermore, CUP model simulations were extended to investigate the influence of operating and intrinsic parameters on the yield and productivity, and to compare the separation performances of EECCC and DM CCC in various conditions. The results demonstrate that low distribution constants, usually a KD less than 1, and a selectivity > 1.3, under specific flowrate ranges, increase both productivity and yield. From these results, generalized optimization and scaleup guidelines are proposed that can apply to research settings and to industrial processes to maximize preparative CCC performance.

BIOMASS FUELS,INORGANIC, ORGANIC, PHYSICAL, AND AN↗

Understanding and predicting soot generation in turbulent non-premixed jet flames.

This report documents the results of a project funded by DoD's Strategic Environmental Research and Development Program (SERDP) on the science behind development of predictive models for soot emission from gas turbine engines. Measurements of soot formation were performed in laminar flat premixed flames and turbulent non-premixed jet flames at 1 atm pressure and in turbulent liquid spray flames under representative conditions for takeoff in a gas turbine engine. The laminar flames and open jet flames used both ethylene and a prevaporized JP-8 surrogate fuel composed of n-dodecane and m-xylene. The pressurized turbulent jet flame measurements used the JP-8 surrogate fuel and compared its combustion and sooting characteristics to a world-average JP-8 fuel sample. The pressurized jet flame measurements demonstrated that the surrogate was representative of JP-8, with a somewhat higher tendency to soot formation. The premixed flame measurements revealed that flame temperature has a strong impact on the rate of soot nucleation and particle coagulation, but little sensitivity in the overall trends was found with different fuels. An extensive array of non-intrusive optical and laser-based measurements was performed in turbulent non-premixed jet flames established on specially designed piloted burners. Soot concentration data was collected throughout the flames, together with instantaneous images showing the relationship between soot and the OH radical and soot and PAH. A detailed chemical kinetic mechanism for ethylene combustion, including fuel-rich chemistry and benzene formation steps, was compiled, validated, and reduced. The reduced ethylene mechanism was incorporated into a high-fidelity LES code, together with a moment-based soot model and models for thermal radiation, to evaluate the ability of the chemistry and soot models to predict soot formation in the jet diffusion flame. The LES results highlight the importance of including an optically-thick radiation model to accurately predict gas temperatures and thus soot formation rates. When including such a radiation model, the LES model predicts mean soot concentrations within 30% in the ethylene jet flame.

33 ADVANCED PROPULSION SYSTEMS↗

Designing accurate emulators for scientific processes using calibration-driven deep models

Abstract Predictive models that accurately emulate complex scientific processes can achieve speed-ups over numerical simulators or experiments and at the same time provide surrogates for improving the subsequent analysis. Consequently, there is a recent surge in utilizing modern machine learning methods to build data-driven emulators. In this work, we study an often overlooked, yet important, problem of choosing loss functions while designing such emulators. Popular choices such as the mean squared error or the mean absolute error are based on a symmetric noise assumption and can be unsuitable for heterogeneous data or asymmetric noise distributions. We propose Learn-by-Calibrating, a novel deep learning approach based on interval calibration for designing emulators that can effectively recover the inherent noise structure without any explicit priors. Using a large suite of use-cases, we demonstrate the efficacy of our approach in providing high-quality emulators, when compared to widely-adopted loss function choices, even in small-data regimes.

97 MATHEMATICS AND COMPUTING↗

An Initial Vibro-Acoustic Model for Predicting Electric Motor Noise

A system-level tool to predict electric motor noise, suitable for trade studies of eVTOL and UAM aircraft, does not exist. This work describes the ongoing effort to develop such a tool. It utilizes a 3 phase prediction routine that models the electro-magnetic interactions of the electric motor, the resulting structural vibrations of the motor rotor, and finally the acoustic propagation generated by those structural vibrations.

Electric Motor Noise↗

Search for new phenomena in events with three or more charged leptons in pp collisions at $ \sqrt{s}=8 $ TeV with the ATLAS detector

A generic search for anomalous production of events with at least three charged leptons is presented. The data sample consists of pp collisions at √s=8 TeV collected in 2012 by the ATLAS experiment at the CERN Large Hadron Collider, and corresponds to an integrated luminosity of 20.3 fb -1 . Events are required to have at least three selected lepton candidates, at least two of which must be electrons or muons, while the third may be a hadronically decaying tau. Selected events are categorized based on their lepton flavour content and signal regions are constructed using several kinematic variables of interest. No significant deviations from Standard Model predictions are observed. Model-independent upper limits on contributions from beyond the Standard Model phenomena are provided for each signal region, along with prescription to re-interpret the limits for any model. Constraints are also placed on models predicting doubly charged Higgs bosons and excited leptons. For doubly charged Higgs bosons decaying to eτ or μτ, lower limits on the mass are set at 400 GeV at 95% confidence level. For excited leptons, constraints are provided as functions of both the mass of the excited state and the compositeness scale Λ, with the strongest mass constraints arising in regions where the mass equals Λ. In such scenarios, lower mass limits are set at 3.0 TeV for excited electrons and muons, 2.5 TeV for excited taus, and 1.6 TeV for every excited-neutrino flavour.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Advanced Models and Controls for Prediction and Extension of Battery Lifetime (Presentation)

Predictive models of capacity and power fade must consider a multiplicity of degradation modes experienced by Li-ion batteries in the automotive environment. Lacking accurate models and tests, lifetime uncertainty must presently be absorbed by overdesign and excess warranty costs. To reduce these costs and extend life, degradation models are under development that predict lifetime more accurately and with less test data. The lifetime models provide engineering feedback for cell, pack and system designs and are being incorporated into real-time control strategies.

25 ENERGY STORAGE↗

Global parameterization and validation of a two-leaf light use efficiency model for predicting gross primary production across FLUXNET sites: TL-LUE Parameterization and Validation

We present the first extended validation of satellitemicrowave (MW) liquidwater path (LWP) for low nonprecipitating clouds, from four operational sensors, against ship-borne observations from a three-channel MW radiometer collected along ship transects over the northeast Pacific during May–August 2013. Satellite MW retrievals have an overall correlation of 0.84 with ship observations and a bias of 9.3 g/m2. The bias for broken cloud scenes increases linearly with water vapor path and remains below 17.7 g/m2. In contrast, satelliteMWLWP is unbiased in overcast scenes with correlations up to 0.91, demonstrating that the retrievals are accurate and reliable under these conditions. Satellite MW retrievals produce a diurnal cycle amplitude consistent with ship-based observations (33 g/m2). Observations taken aboard extended ship cruises to evaluate not only satellite MW LWP but also LWP derived from visible/infrared sensors offer a new way to validate this important property over vast oceanic regions.

Zhou, Yanlian↗

Need A Boost? A Comparison of Traditional Commuting Models with the XGBoost Model for Predicting Commuting Flows

Commuting models estimate the number of commuting trips from home to work locations in a given area. Since their infancy, they have been increasingly used in a variety of fields to reduce traffic and pollution, drive infrastructure choices, and solve a variety of other problems. Traditional commuting models, such as gravity and radiation models, typically have a strict structural form and limited number of input variables, which may limit their ability to predict commuting flows as well as machine learning models that might better capture the complex dynamics of the commuting process. To determine whether machine learning models might add value to the field of commuter flow prediction, we compare and discuss the performance of two standard traditional models with the XGBoost machine learning algorithm for predicting home to work commuter flows from a well-known United States commuting dataset. We find that the XGBoost model outperforms the traditional models on three commonly used metrics, indicating that machine learning models may add value to the field of commuter flow prediction.

Morton, April↗

MSL MMRTG Power Modeling for Lifetime Performance Predictions

The Multi-Mission Radioisotope Thermoelectric Generator (MMRTG) has been providing power to the Mars Science Laboratory (MSL) rover, Curiosity, for surface operations since 2012. The Jet Propulsion Laboratory (JPL) has employed the Life Performance Prediction Model (LPPM) to generate predictions and outputs based upon flight data from the rover and experimental data from the thermoelectric (TE) couples. In order to make these predictions, LPPM requires various spacecraft inputs relevant to RTG performance such as fin root temperature, load voltage, general purpose heat source (GPHS) inventory, etc., combined with couple data such as thermoelectric properties, interface degradation, and sublimation properties. With tens of thousands of hours of flight data to date, LPPM can not only track performance to date, but also make predictions on the power output of the MMRTG through the 17-year end-of-design-life (EODL). These predictions are useful for mission planners and surface operators who rely on the MMRTG for power in order to perform the critical science necessary during the mission. This paper will showcase the power data extracted from the flight electronics, comparing it to the predictions and outputs generated from LPPM.

Pinkowski, Stanley↗