Search NASA⌕ Search

SEARCH · Search NASA

Results for “predict”

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 1,549 records · Page 86

Uncertainty Aware Deep Learning for Fault Prediction Using Multivariate Time Series Signals

The superconducting radio-frequency cavities are a crucial component of the Continuous Electron Beam Accelerator Facility (CEBAF) at Jefferson Lab. When a cavity faults, beam delivery to experimental end users is disrupted. Prediction of cavity faults prior to onset is essential to reduce operation and maintenance costs. In this work, a parallel long short-term memory (LSTM)-convolution neural network (CNN)-based deep learning (DL) model is proposed to predict impending faults using pre-fault signals. Further, we introduce an uncertainty quantification approach using Monte Carlo dropout with the LSTM-CNN model to ascertain confidence in the prediction. The model was tested using multivariate time series signals from stable cavity operations and before faults. Initial results show that on the test dataset, the model can identify impending faults before their onset with an average 10-fold cross validation accuracy of 97.39% and a standard deviation of 0.12% using a 100-ms time window. It is also observed that the model performs better as the prediction time moves closer to the fault onset. For additional context, we compare the performance of the model with three machine-learning-based (ML) fault prediction models. Our proposed parallel LSTM-CNN-based DL method shows better performance than the ML-based methods.

Rahman, Md Monibor↗

Generalized and Mechanistic PV Module Performance Prediction From Computer Vision and Machine Learning on Electroluminescence Images

Electroluminescence (EL) imaging of photovoltiac (PV) modules offers high-speed, high-resolution information about device performance, affording opportunities for greater insight and efficiency in module characterization across manufacturing, research and development, and power plant operations and management. Predicting module electrical properties from EL image features is a critical step toward these applications. In this article, we demonstrate quantification of both generalized and performance mechanism-specific EL image features, using pixel intensity-based and machine learning classification algorithms. From EL image features, we build predictive models for PV module power and series resistance, using time-series current-voltage (I-V) and EL data obtained stepwise on five brands of modules spanning three Si cell types through two accelerated exposures: damp heat (DH) (85°C/85% RH) and thermal cycling (TC) (IEC 61215). Overall, 195 pairs of EL images and I-V characteristics were analyzed, yielding 11700 individual PV cell images. A convolutional neural network was built to classify cells by the severity of busbar corrosion with high accuracy (95%). Generalized power predictive models estimated the maximum power of PV modules from EL images with high confidence and an adjusted-R 2 of 0.88, across all module brands and cell types in extended DH and TC exposures. Mechanistic degradation prediction was demonstrated by quantification of busbar corrosion in EL images of three module brands in DH, and subsequent modeling of series resistance using these mechanism-specific EL image features. For modules exhibiting busbar corrosion, we demonstrated series resistance predictive models with adjusted-R 2 of up to 0.73.

14 SOLAR ENERGY↗

Advanced Photovoltaic Module Characterization: Using Image Transformers for Current–Voltage Curve Prediction From Electroluminescence Images

Individual photovoltaic (PV) module health monitoring can be a daunting task for operation and maintenance of solar farms. Modules can be inspected through luminescence, thermal imaging, and current–voltage (I–V) curve analyzes for identification of damage and power loss. I–V curves provide easily interpretable data to determine module health as they directly provide electrical performance metrics. However, in order to obtain these curves, modules must be disconnected from the array and either removed to a solar simulator or characterized in situ with corrections for module temperature, the incident solar spectrum, and intensity. Luminescence or thermal images of a module are relatively easy to acquire in situ. Electroluminescence (EL) images highlight physical defects in the modules but do not provide easily interpretable features to correlate with electrical performance. This work presents a SWin transformer network to predict I–V curves for PV modules from their corresponding EL images. The predicted I–V curves allow the accurate prediction of the maximum power point (MPP), short-circuit current I sc , and open-circuit voltage V oc with a mean error less of than 1%. Comparing single diode model (SDM) parameters extracted from the predicted curves to those extracted from the true curves, the series resistance R s demonstrates a mean error of 5.19%, and the photocurrent I a mean error of 0.197%. The shunt resistance R sh and dark current Io parameters are predicted with larger errors because of their sensitivity to small changes in the I–V curve.

Byford, Brandon K. [New Mexico State Univ., Las Cr↗

Invariant Features for Accurate Predictions of Quantum Chemical UV-vis Spectra of Organic Molecules

Including invariance of global properties of a phys-ical system as an intrinsic feature in graph neural networks (GNNs) enhances the model's robustness and generalizability and reduces the amount of training data required to obtain a desired accuracy for predictions of these properties. Existing open source GNN libraries construct invariant features only for specific GNN architectures. This precludes the generalization of invariant features to arbitrary message passing neural network (MPNN) layers which, in turn, precludes the use of these libraries for new, user-specified predictive tasks. To address this limitation, we implement invariant MPNNs into the flexible and scalable HydraGNN architecture. HydraGNN enables a seamless switch between various MPNNs in a unified layer sequence and allows for a fair comparison between the predictive performance of different MPNNs. We trained this enhanced HydraGNN archi-tecture on the ultraviolet-visible (UV-vis) spectrum of GDB-9 molecules, a feature that describes the molecule's electronic exci-tation modes, computed with time-dependent density functional tight binding (TD-DFTB) and available open source through the GDB-9-Ex dataset. We assess the robustness (i.e., accuracy and generalizability) of the predictions obtained using different invariant MPNNs with respect to different values of the full width at half maximum (FWHM) for the Gaussian smearing of the theoretical peaks. Our numerical results show that incorporating invariance in the HydraGNN architecture significantly enhances both accuracy and generalizability in predicting UV-vis spectra of organic molecules.

Baker, Justin↗

Impedance-Based Prediction of Distortions Generated by Resonance in Grid-Connected Converters

Small-signal impedance-based analysis can effectively predict the frequency and damping of resonance modes in power electronic systems. However, it cannot predict the magnitude of resonance or explain resonance-generated distortions in the absence of an external disturbance at the resonance frequency. This paper presents an impedance-based theory for the prediction of the magnitude of resonance or resonance-generated distortions in grid-connected converters. It is discovered that the impedance response of a converter starts changing with the magnitude of resonance at its terminals. The changing converter impedance response may stabilize an unstable growing resonance mode beyond a certain magnitude, at which point the converter enters a limit cycle mode of sustained oscillations. The proposed theory uses large-signal impedance for the prediction of resonance-generated distortions; the large-signal impedance of an electrical equipment represents its impedance response for different magnitudes of perturbation injected at its terminals. Large-signal impedance-based prediction of resonance-generated distortions is demonstrated for a three-phase grid-connected voltage-source converter.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Structure-Informed Graph Learning of Networked Dependencies for Online Prediction of Power System Transient Dynamics

Online transient analysis plays an increasingly important role in dynamic power grids as the renewable generation continues growing. Traditional numerical methods for transient analysis not only are computationally intensive but also require precise contingency information as input, and therefore, are not suitable for online applications. Existing online transient assessment studies focus on the determination of post-contingency system stability or stability margin. Here, this paper develops a novel graph-learning framework, Deep-learning Neural Representation or DNR, for online prediction, of the time-series trajectories of the system states using initial system responses that can be measured by phasor measurement units (PMUs). The proposed DNR framework consists of two sequential modules: a Network Constructor that captures network dependencies among generators, and a Dynamics Predictor that predicts the system trajectories. The key to improved prediction performance is the introduction of the spatio-temporal message-passing operations into graph neural networks with structural knowledge. Its effectiveness and scalability are validated through comparative studies, demonstrating the prediction performance under different contingency scenarios for systems of different sizes. This framework provides a solution to online predicting post-fault system dynamics based on real-time PMU measurements. Additionally, it can also be applied to facilitate the offline transient simulation without simulating the entire trajectories.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Rapid Prediction of Low (<1%) trans Fat Content in Edible Oils and Fast Food Lipid Extracts by Infrared Spectroscopy and Partial Least Squares Regression

Abstract The United States Food and Drug Administration (FDA) ruled that partially hydrogenated oils (PHO), the major dietary source of industrially produced trans fat (TF), were no longer “generally recognized as safe (GRAS)” for any use in human food. Consequently, the objective of this study was to develop a rapid screening procedure using attenuated total reflection Fourier‐transform infrared (ATR‐FTIR) spectroscopy in conjunction with partial least squares regression (PLSR) for the quantitative and accurate prediction of low concentrations of trans fatty acid (TFAs) (<1% of total fatty acids (FAs)). Broad‐based calibration models were developed for a combined set of samples consisting of edible oils and fast food lipid extracts. Predicted concentrations of TFAs in the two matrices showed good correlation with the primary reference data generated by gas chromatography (GC) ( R 2 > 0.99) and high accuracy as evidenced by low root‐mean‐square error of cross‐validation (RMSECV) values. The lowest TFA concentration, determined by GC to be 0.13% of total FAs, was accurately predicted by ATR‐FTIR/PLSR as 0.18% of total FAs. This simple, rapid ATR‐FTIR/PLSR methodology has the potential for use as a screening alternative to conventional gas chromatographic methods for predicting the TFA content of edible oils and food lipid extracts for regulatory purposes and quality control of raw material and processed food. Practical Applications F DA ruled that partially hydrogenated oils were no longer “generally recognized as safe (GRAS)” for any use in human food. Consequently, we have proposed a rapid screening procedure, based on infrared spectroscopy and chemometrics, to rapidly and accurately predict low concentrations of trans fatty acids (<1% of total fatty acids) in edible oils and food lipid extracts.

Karunathilaka, Sanjeewa R.↗

Predicting invasion in grassland ecosystems: is exotic dominance the real embarrassment of richness?

Invasions have increased the size of regional species pools, but are typically assumed to reduce native diversity. However, global-scale tests of this assumption have been elusive because of the focus on exotic species richness, rather than relative abundance. This is problematic because low invader richness can indicate invasion resistance by the native community or, alternatively, dominance by a single exotic species. Here, we used a globally replicated study to quantify relationships between exotic richness and abundance in grass-dominated ecosystems in 13 countries on six continents, ranging from salt marshes to alpine tundra. We tested effects of human land use, native community diversity, herbivore pressure, and nutrient limitation on exotic plant dominance. Despite its widespread use, exotic richness was a poor proxy for exotic dominance at low exotic richness, because sites that contained few exotic species ranged from relatively pristine (low exotic richness and cover) to almost completely exotic-dominated ones (low exotic richness but high exotic cover). Both exotic cover and richness were predicted by native plant diversity (native grass richness) and land use (distance to cultivation). Although climate was important for predicting both exotic cover and richness, climatic factors predicting cover (precipitation variability) differed from those predicting richness (maximum temperature and mean temperature in the wettest quarter). Herbivory and nutrient limitation did not predict exotic richness or cover. Exotic dominance was greatest in areas with low native grass richness at the site- or regional-scale. Although this could reflect native grass displacement, a lack of biotic resistance is a more likely explanation, given that grasses comprise the most aggressive invaders. These findings underscore the need to move beyond richness as a surrogate for the extent of invasion, because this metric confounds monodominance with invasion resistance. Monitoring species’ relative abundance will more rapidly advance our understanding of invasions.

60 APPLIED LIFE SCIENCES↗

Microbial models with data‐driven parameters predict stronger soil carbon responses to climate change

Abstract Long‐term carbon (C) cycle feedbacks to climate depend on the future dynamics of soil organic carbon ( SOC ). Current models show low predictive accuracy at simulating contemporary SOC pools, which can be improved through parameter estimation. However, major uncertainty remains in global soil responses to climate change, particularly uncertainty in how the activity of soil microbial communities will respond. To date, the role of microbes in SOC dynamics has been implicitly described by decay rate constants in most conventional global carbon cycle models. Explicitly including microbial biomass dynamics into C cycle model formulations has shown potential to improve model predictive performance when assessed against global SOC databases. This study aimed to data‐constrained parameters of two soil microbial models, evaluate the improvements in performance of those calibrated models in predicting contemporary carbon stocks, and compare the SOC responses to climate change and their uncertainties between microbial and conventional models. Microbial models with calibrated parameters explained 51% of variability in the observed total SOC , whereas a calibrated conventional model explained 41%. The microbial models, when forced with climate and soil carbon input predictions from the 5th Coupled Model Intercomparison Project ( CMIP 5), produced stronger soil C responses to 95 years of climate change than any of the 11 CMIP 5 models. The calibrated microbial models predicted between 8% (2‐pool model) and 11% (4‐pool model) soil C losses compared with CMIP 5 model projections which ranged from a 7% loss to a 22.6% gain. Lastly, we observed unrealistic oscillatory SOC dynamics in the 2‐pool microbial model. The 4‐pool model also produced oscillations, but they were less prominent and could be avoided, depending on the parameter values.

Hararuk, Oleksandra↗

Predicting soil carbon changes in switchgrass grown on marginal lands under climate change and adaptation strategies

Abstract The United States Great Lakes Region (USGLR) is a critical geographic area for future bioenergy production. Switchgrass ( Panicum virgatum ) is widely considered a carbon (C)‐neutral or C‐negative bioenergy production system, but projected increases in air temperature and precipitation due to climate change might substantially alter soil organic C (SOC) dynamics and storage in soils. This study examined long‐term SOC changes in switchgrass grown on marginal land in the USGLR under current and projected climate, predicted using a process‐based model (Systems Approach to Land‐Use Sustainability) extensively calibrated with a wealth of plant and soil measurements at nine experimental sites. Simulations indicate that these soils are likely a net C sink under switchgrass (average gain 0.87 Mg C ha −1 year −1 ), although substantial variation in the rate of SOC accumulation was predicted (range: 0.2–1.3 Mg C ha −1 year −1 ). Principal component analysis revealed that the predicted intersite variability in SOC sequestration was related in part to differences in climatic characteristics, and to a lesser extent, to heterogeneous soils. Although climate change impacts on switchgrass plant growth were predicted to be small (4%–6% decrease on average), the increased soil respiration was predicted to partially negate SOC accumulations down to 70% below historical rates in the most extreme scenarios. Increasing N fertilizer rate and decreasing harvest intensity both had modest SOC sequestration benefits under projected climate, whereas introducing genotypes better adapted to the longer growing seasons was a much more effective strategy. Best‐performing adaptation scenarios were able to offset >60% of the climate change impacts, leading to SOC sequestration 0.7 Mg C ha −1 year −1 under projected climate. On average, this was 0.3 Mg C ha −1 year −1 more C sequestered than the no adaptation baseline. These findings provide crucial knowledge needed to guide policy and operational management for maximizing SOC sequestration of future bioenergy production on marginal lands in the USGLR.

09 BIOMASS FUELS↗

The Predictive Power of Electronic Polarizability for Tailoring the Refractivity of High Index Glasses Optical Basicity Versus the Single Oscillator Model

Four compositions of high density (~8 g/cm3) heavy metal oxide glasses composed of PbO, Bi2O3, and Ga2O3 were produced and refractivity parameters (refractive index and density) were computed and measured. Optical basicity was computed using three different models – average electronegativity, ionic-covalent parameter, and energy gap – and the basicity results were used to compute oxygen polarizability and subsequently refractive index. Refractive indices were measured in the visible and infrared at 0.633 μm, 1.55 μm, 3.39 μm, 5.35 μm, 9.29 μm, and 10.59 μm using a unique prism coupler setup, and data were fitted to the Sellmeier expression to obtain an equation of the dispersion of refractive index with wavelength. Using this dispersion relation, single oscillator energy, dispersion energy, and lattice energy were determined. Oscillator parameters were also calculated for the various glasses from their oxide values as an additional means of predicting index. Calculated dispersion parameters from oxides underestimate the index by 3 to 4%. Predicted glass index from optical basicity, based on component oxide energy gaps, underpredicts the index at 0.633 μm by only 2%, while other basicity scales are less accurate. The predicted energy gap of the glasses based on this optical basicity overpredicts the Tauc optical gap as determined by transmission measurements by 6 to 10%. These results show that for this system, density, refractive index in the visible, and energy gap can be reasonably predicted using only composition, optical basicity values for the constituent oxides, and partial molar volume coefficients. Calculations such as these are useful for a priori prediction of optical properties of glasses.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A Predictive Model for Survival of Escherichia coli O157:H7 and Generic E. coli in Soil Amended with Untreated Animal Manure

Abstract This study aimed at developing a predictive model that captures the influences of a variety of agricultural and environmental variables and is able to predict the concentrations of enteric bacteria in soil amended with untreated Biological Soil Amendments of Animal Origin (BSAAO) under dynamic conditions. We developed and validated a Random Forest model using data from a longitudinal field study conducted in mid‐Atlantic United States investigating the survival of Escherichia coli O157:H7 and generic E. coli in soils amended with untreated dairy manure, horse manure, or poultry litter. Amendment type, days of rain since the previous sampling day, and soil moisture content were identified as the most influential agricultural and environmental variables impacting concentrations of viable E. coli O157:H7 and generic E. coli recovered from amended soils. Our model results also indicated that E. coli O157:H7 and generic E. coli declined at similar rates in amended soils under dynamic field conditions.The Random Forest model accurately predicted changes in viable E. coli concentrations over time under different agricultural and environmental conditions. Our model also accurately characterized the variability of E. coli concentration in amended soil over time by providing upper and lower prediction bound estimates. Cross‐validation results indicated that our model can be potentially generalized to other geographic regions and incorporated into a risk assessment for evaluating the risks associated with application of untreated BSAAO. Our model can be validated for other regions and predictive performance also can be enhanced when data sets from additional geographic regions become available.

Pang, Hao↗

Numerical Prediction of Cyclic Variability in a Spark Ignition Engine Using a Parallel Large Eddy Simulation Approach

Cycle-to-cycle variability (CCV) is detrimental to IC engine operation and can lead to partial burn, misfire, and knock. Predicting CCV numerically is extremely challenging due to two key reasons. First, high-fidelity methods such as large eddy simulation (LES) are required to accurately resolve the in-cylinder turbulent flow field both spatially and temporally. Second, CCV is experienced over long timescales and hence the simulations need to be performed for hundreds of consecutive cycles. Ameen et al. (2017, "Parallel Methodology to Capture Cyclic Variability in Motored Engines,"Int. J. Engine Res., 18(4), pp. 366-377.) developed a parallel perturbation model (PPM) approach to dissociate this long time-scale problem into several shorter time-scale problems. This strategy was demonstrated for motored engine and it was shown that the mean and variance of the in-cylinder flow field was captured reasonably well by this approach. In the present study, this PPM approach is extended to simulate the CCV in a fired port-fuel injected (PFI) spark ignition (SI) engine. The predictions from this approach are also shown to be similar to the consecutive LES cycles. It is shown that the parallel approach is able to predict the coefficient of variation (COV) of the in-cylinder pressure and burn raterelated parameters with sufficient accuracy, and is also able to predict the qualitative trends in CCV with changing operating conditions. It is shown that this new approach is able to give accurate predictions of the CCV in fired engines in less than one-tenth of the time required for the conventional approach of simulating consecutive engine cycles.

Ameen, Muhsin M.↗

An Assessment of Homogeneous Mixture Method Cavitation Models in Predicting Cavitation in Nozzle Flow

The homogeneous mixture method (HMM) is a popular class of models used in the computational prediction of cavitation. Several cavitation models have been developed for use with HMM to govern the development and destruction of vapor in a fluid system. Two models credited to Kunz and Schnerr–Sauer are studied in this paper. The goal of this work is to provide an assessment of the two cavitation submodels in their ability to predict cavitation in nozzle flow. Validation data were obtained via experiments which employ both passive cavitation detection, (PCD) via acoustic sensing and optical cavitation detection (OCD) via camera imaging. The experiments provide quantitative information on cavitation inception and qualitative information on the vapor in the nozzle. The results show that initial vapor formation is not predicted precisely but within reason. A sensitivity analysis of the models to input parameters shows that the Schnerr–Sauer method does not depend upon the estimation of nuclei size and number density. Small changes in the vapor formation rate but not the total vapor volume can be seen when weighting parameters are modified. In contrast, changes to the input parameters for the Kunz model greatly change the final total vapor volume prediction. The assessment also highlights the influence of vapor convection within the method. Finally, the analysis shows that if the fluid and nozzle walls do not support nuclei larger than 40 μm, the methods would still predict cavitation when indeed there would be none in practice.

Engineering↗

A Framework for Inverse Prediction Using Functional Response Data

Inverse prediction models have commonly been developed to handle scalar data from physical experiments. However, it is not uncommon for data to be collected in functional form. When data are collected in functional form, it must be aggregated to fit the form of traditional methods, which often results in a loss of information. For expensive experiments, this loss of information can be costly. In this study, we introduce the functional inverse prediction (FIP) framework, a general approach which uses the full information in functional response data to provide inverse predictions with probabilistic prediction uncertainties obtained with the bootstrap. The FIP framework is a general methodology that can be modified by practitioners to accommodate many different applications and types of data. We demonstrate the framework, highlighting points of flexibility, with a simulation example and applications to weather data and to nuclear forensics. Results show how functional models can improve the accuracy and precision of predictions.

42 ENGINEERING↗

Predicting the Cetane Number, Yield Sooting Index, Kinematic Viscosity, and Cloud Point for Catalytically Upgraded Pyrolysis Oil Using Artificial Neural Networks

Abstract The conversion of biomass using fast pyrolysis has the potential to be significantly less expensive at scale compared to alternative methods such as fermentation and gasification. Selective upgrading of the products of fast pyrolysis through chemical catalysis produces compounds with lower oxygen content and lower acidity; however, identifying the specific catalytic pathways for producing viable fuels and fuel additives often requires a trial-and-error approach. Specifically, key properties of the compounds must be experimentally tested to evaluate the viability of the resultant compounds. The present work proposes predictive models constructed with artificial neural networks (ANNs) for cetane number (CN), yield sooting index (YSI), kinematic viscosity (KV), and cloud point (CP), with blind test set median absolute errors of 5.14 cetane units, 3.36 yield sooting index units, 0.07 millimeters squared per second, and 4.89 degrees Celsius, respectively. Furthermore, the cetane number, yield sooting index, kinematic viscosity, and cloud point were predicted for over three hundred expected products from the catalytic upgrading of pyrolysis oil. It was discovered that 130 of these compounds have predicted cetane numbers greater than 40, with four of these compounds possessing predicted yield sooting index values significantly less than that of diesel fuel and predicted viscosities and cloud points comparable to that of diesel fuel.

09 BIOMASS FUELS↗

Predicting the Cetane Number, Sooting Tendency, and Energy Density of Terpene Fuel Additives

Abstract Discovering renewable fuels and fuel additives is paramount in reducing carbon emissions from internal combustion engines. Terpenes, a group of compounds that can be synthesized from plant matter and microorganisms, have gained significant interest in recent years as promising candidates for fuels/additives. Terpenes are a diverse class of compounds that contain rings and methyl branches, resulting in high energy densities and optimal cold weather behavior. Their variation in bond order, carbon chains, and functional groups lead to varying degrees of soot formation and performance in existing engines. The present work leverages predictive models, namely artificial neural networks, to predict the cetane number (CN), sooting tendency (quantified with yield sooting index, YSI), and energy density (quantified with lower heating value, LHV) of terpenes and hydrogenated terpenes whose sooting propensities were previously determined through experimental means. Predicted sooting propensities of these terpenes are compared with experimental values, and predicted cetane numbers and energy densities are used to comment on the compounds’ ability to act as fuels/additives. Expected prediction errors for CN, YSI, and LHV, defined by blind test set median absolute error, are within 5.56 cetane units, 3.63 yield sooting index units, and 0.77 MJ/kg respectively. Additionally, the present work investigates a variety of correlation/dependence metrics for property-property relationships, furthering our understanding of how combustion-relevant properties are related.

09 BIOMASS FUELS↗

Predicting the fidelity of JPEG2000 compressed CT images using DICOM header information

Purpose: To propose multiple logistic regression (MLR) and artificial neural network (ANN) models constructed using digital imaging and communications in medicine (DICOM) header information in predicting the fidelity of Joint Photographic Experts Group (JPEG) 2000 compressed abdomen computed tomography (CT) images. Methods: Our institutional review board approved this study and waived informed patient consent. Using a JPEG2000 algorithm, 360 abdomen CT images were compressed reversibly (n = 48, as negative control) or irreversibly (n = 312) to one of different compression ratios (CRs) ranging from 4:1 to 10:1. Five radiologists independently determined whether the original and compressed images were distinguishable or indistinguishable. The 312 irreversibly compressed images were divided randomly into training (n = 156) and testing (n = 156) sets. The MLR and ANN models were constructed regarding the DICOM header information as independent variables and the pooled radiologists' responses as dependent variable. As independent variables, we selected the CR (DICOM tag number: 0028, 2112), effective tube current-time product (0018, 9332), section thickness (0018, 0050), and field of view (0018, 0090) among the DICOM tags. Using the training set, an optimal subset of independent variables was determined by backward stepwise selection in a four-fold cross-validation scheme. The MLR and ANN models were constructed with the determined independent variables using the training set. The models were then evaluated on the testing set by using receiver-operating-characteristic (ROC) analysis regarding the radiologists' pooled responses as the reference standard and by measuring Spearman rank correlation between the model prediction and the number of radiologists who rated the two images as distinguishable. Results: The CR and section thickness were determined as the optimal independent variables. The areas under the ROC curve for the MLR and ANN predictions were 0.91 (95% CI; 0.86, 0.95) and 0.92 (0.87, 0.96), respectively. The correlation coefficients of the MLR and ANN predictions with the number of radiologists who responded as distinguishable were 0.76 (0.69, 0.82, p < 0.001) and 0.78 (0.71, 0.83, p < 0.001), respectively. Conclusions: The MLR and ANN models constructed using the DICOM header information offer promise in predicting the fidelity of JPEG2000 compressed abdomen CT images.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗