Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “robust regression”

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 145 records · Page 8

Chemical studies of H chondrites. 4: New data and comparison of Antarctic suites

We report data for the trace elements Au, Co, Sb, Ga, Rb, Ag, Se, Cs, Te, Zn, Cd, Bi, Ti, and In (ordered by putative volatility during nebular condensation and accretion) determined by neutron activation analysis in 13 H5 chondrites from Victoria Land and 20 H4-6 chondrites from Queen Maud Land, Antarctica. These and earlier results provide Antarctic sample suites of 34 chondrites from Victoria Land and 25 from Queen Maud Land. Treatment of data for the most volatile 10 elements (Rb to In) in these studies by multivariate statistical techniques more robust, as well as more conservative, than conventional linear discriminant analysis and logistic regression demonstrates that compositions differ at marginally significant levels. This difference cannot be explained by trivial (terrestrial) causes and becomes more significant, despite the smaller size of the database, when comparisons are limited to data from a single analyst and when all upper limits are eliminated from consideration. The Victoria Land and Queen Maud Land suites have different mean terrestrial ages (approximately 300 kyr and approximately 100 kyr, respectively) and age distributions, suggesting that a time-dependent variation of chondritic sources with different thermal histories is responsible. As a result, these two Antarctic suites are, on average, chemically distinguishable from each other. Since H chondrites serve as a paradigm for other meteorite classes, these results indicate that the near-Earth populations of planetary materials varied with time on the 10(exp 5)-year timescale.

Wolf, Stephen F.↗

Ames Hybrid Combustion Facility

The report summarizes the design, fabrication, safety features, environmental impact, and operation of the Ames Hybrid-Fuel Combustion Facility (HCF). The facility is used in conducting research into the scalability and combustion processes of advanced paraffin-based hybrid fuels for the purpose of assessing their applicability to practical rocket systems. The facility was designed to deliver gaseous oxygen at rates between 0.5 and 16.0 kg/sec to a combustion chamber operating at pressures ranging from 300 to 900. The required run times were of the order of 10 to 20 sec. The facility proved to be robust and reliable and has been used to generate a database of regression-rate measurements of paraffin at oxygen mass flux levels comparable to those of moderate-sized hybrid rocket motors.

Zilliac, Greg↗

A diagnostic analysis of the VVP single-doppler retrieval technique

A diagnostic analysis of the VVP (volume velocity processing) retrieval method is presented, with emphasis on understanding the technique as a linear, multivariate regression. Similarities and differences to the velocity-azimuth display and extended velocity-azimuth display retrieval techniques are discussed, using this framework. Conventional regression diagnostics are then employed to quantitatively determine situations in which the VVP technique is likely to fail. An algorithm for preparation and analysis of a robust VVP retrieval is developed and applied to synthetic and actual datasets with high temporal and spatial resolution. A fundamental (but quantifiable) limitation to some forms of VVP analysis is inadequate sampling dispersion in the n space of the multivariate regression, manifest as a collinearity between the basis functions of some fitted parameters. Such collinearity may be present either in the definition of these basis functions or in their realization in a given sampling configuration. This nonorthogonality may cause numerical instability, variance inflation (decrease in robustness), and increased sensitivity to bias from neglected wind components. It is shown that these effects prevent the application of VVP to small azimuthal sectors of data. The behavior of the VVP regression is further diagnosed over a wide range of sampling constraints, and reasonable sector limits are established.

Boccippio, Dennis J.↗

Combining Emergent Constraints for Climate Sensitivity

A method is proposed for combining information from several emergent constraints into a probabilistic estimate for a climate sensitivity proxy Y such as equilibrium climate sensitivity (ECS). The method is based on fitting a multivariate Gaussian PDF for Y and the emergent constraints using an ensemble of global climate models (GCMs); it can be viewed as a form of multiple linear regression of Y on the constraints. The method accounts for uncertainties in sampling this multidimensional PDF with a small number of models, for observational uncertainties in the constraints, and for overconfidence about the correlation of the constraints with the climate sensitivity. Its general form (Method C) accounts for correlations between the constraints. Method C becomes less robust when some constraints are too strongly related to each other; this can be mitigated using regularization approaches such as ridge regression. An illuminating special case, Method U, neglects any correlations between constraints except through their mutual relationship to the climate proxy; it is more robust to small GCM sample size and is appealingly interpretable. These methods are applied to ECS and the climate feedback parameter using a previously published set of 11 possible emergent constraints derived from climate models in the Coupled Model Intercomparison Project (CMIP). The ±2σ posterior range of ECS for Method C with no overconfidence adjustment is 4.3 ± 0.7 K. For Method U with a large overconfidence adjustment, it is 4.0 ± 1.3 K. This study adds confidence to past findings that most constraints predict higher climate sensitivity than the CMIP mean.

54 ENVIRONMENTAL SCIENCES↗

Developing a robust strength model using physically-informed genetic programming

The strength of materials is influenced by a range of external conditions, such as temperature and deformation rate. Consequently, materials that demonstrate substantial variations in their mechanical behavior due to fluctuations in temperature and strain rate require complex strength models to accurately predict material performance in real-world applications. To predict such complex behavior, a robust and flexible strength model is necessary. In this work, we utilize genetic programming-based symbolic regression (GPSR) to develop data-driven strength models that accurately represent the measured stress–strain responses of tin across a wide range of strain, strain rate and temperature regimes. The GPSR models are constrained by physically-informed conditions, which leads to significant improvement in extrapolation. The best model is integrated into a multi-physics code to perform Taylor impact simulations, validating the model’s accuracy and robustness. In conclusion, the model predictions showed excellent agreement with experimental results, particularly when compared to predictions using traditional strength models.

Genetic programming↗

Understanding and Verifying Neural Networks

Deep Neural Networks (DNNs) have gained immense popularity in recent times and have widespread use in applications such as image classification, sentiment analysis, speech recognition and also in safety-critical applications such as autonomous driving. However, they suffer limitations such as lack of explainability and robustness which raise safety and security concerns in their usage. Further, the complex structure and large input spaces of DNNs act as an impediment to thorough verification and testing. The SafeDNN project at the Robust Software Engineering (RSE) group at NASA aims at exploring techniques to ensure that systems that use deep neural networks are safe, robust and interpretable. In this talk, I will be presenting our technique Prophecy that automatically infers formal properties of deep neural network models. The tool extracts patterns based on neuron activations as preconditions that imply certain desirable output properties of the model. I would be highlighting case studies that use Prophecy in obtaining explanations for network decisions, understanding correct and incorrect behavior, providing formal guarantees wrt safety and robustness, and debugging neural network models. We have applied the tool on image classification networks, neural network controllers providing turn advisories in unmanned aircrafts, regression models used for autonomous center-line tracking in aircrafts and neural network object detectors

Deep Neural Networks↗

Automated Algorithms for Quantum-Level Accuracy in Atomistic Simulations: LDRD Final Report.

This report summarizes the result of LDRD project 12-0395, titled "Automated Algorithms for Quantum-level Accuracy in Atomistic Simulations." During the course of this LDRD, we have developed an interatomic potential for solids and liquids called Spectral Neighbor Analysis Poten- tial (SNAP). The SNAP potential has a very general form and uses machine-learning techniques to reproduce the energies, forces, and stress tensors of a large set of small configurations of atoms, which are obtained using high-accuracy quantum electronic structure (QM) calculations. The local environment of each atom is characterized by a set of bispectrum components of the local neighbor density projected on to a basis of hyperspherical harmonics in four dimensions. The SNAP coef- ficients are determined using weighted least-squares linear regression against the full QM training set. This allows the SNAP potential to be fit in a robust, automated manner to large QM data sets using many bispectrum components. The calculation of the bispectrum components and the SNAP potential are implemented in the LAMMPS parallel molecular dynamics code. Global optimization methods in the DAKOTA software package are used to seek out good choices of hyperparameters that define the overall structure of the SNAP potential. FitSnap.py, a Python-based software pack- age interfacing to both LAMMPS and DAKOTA is used to formulate the linear regression problem, solve it, and analyze the accuracy of the resultant SNAP potential. We describe a SNAP potential for tantalum that accurately reproduces a variety of solid and liquid properties. Most significantly, in contrast to existing tantalum potentials, SNAP correctly predicts the Peierls barrier for screw dislocation motion. We also present results from SNAP potentials generated for indium phosphide (InP) and silica (SiO 2 ). We describe efficient algorithms for calculating SNAP forces and energies in molecular dynamics simulations using massively parallel computers and advanced processor ar- chitectures. Finally, we briefly describe the MSM method for efficient calculation of electrostatic interactions on massively parallel computers.

Thompson, Aidan Patrick↗

Modeling longitudinal INMA(1) with COM–Poisson innovation under non-stationarity: application to medical data

This paper introduces an observation-driven (OD) longitudinal integer-valued moving average model of order 1 (INMA(1)) with COM–Poisson innovations under non-stationary moment conditions. This new longitudinal model provides lot of practical flexibility in terms of modeling a wide range of over-, under-dispersion and any mixed level of dispersion. In this set-up, the model parameters of primary interest consist of the regression and dispersion effects while the serial autocorrelation parameters are treated as nuisance. A robust Generalized Quasi-Likelihood approach is formulated to estimate the different set of parameters. The performance of the estimating algorithm is assessed via Monte Carlo experiments under various combination of the serial and dispersion values and is compared with the existing adaptive generalized method of moments. Application to two medical data: epileptic seizures and polyposis counts are also illustrated.

97 MATHEMATICS AND COMPUTING↗

A Markov chain Monte Carlo (MCMC) Bayesian inference approach to analyze apparent activation barriers and reaction orders from microreactor data

Statistical analysis of steady-state catalytic kinetic data is often limited by data sparsity due to the slow pace at which the data is collected. Data sparsity and limitations in statistical analysis make it difficult to differentiate between mechanistic models and catalytic sites. A Bayesian inference tool is reported for catalysis researchers to estimate error in the determination of reaction orders from steady state microreactor data. The benefits of a Bayesian inference approach are discussed, as an alternative to the more common frequentist approach. The approach incorporates prior knowledge of the system and the data collected to form an error estimate on reaction orders. We investigated the effects of three distinct data treatments—individual fitting of trials, pooled analysis, and constrained regression methods—on the precision and uncertainty of reaction order determinations. To assess the robustness of our findings, we conducted sensitivity analyses to evaluate the influence of Bayesian parameters on uncertainty estimation. Additionally, we utilized synthetic data to illustrate how data quality impacts the precision of uncertainty assessments. We show Bayesian analysis can obtain a more precise estimation of error with a sparse data set than a frequentist analysis. Finally, this work provides strong evidence that the adoption of Bayesian analysis of kinetic data may help researchers make more precise arguments as to the strength of their evidence for a particular mechanistic hypothesis, or in comparing across different catalysts.

42 ENGINEERING↗

Improving E3SM Land Model Photosynthesis Parameterization via Satellite SIF, Machine Learning, and Surrogate Modeling

The parameterization of key photosynthesis parameters is one of the key uncertain sources in modeling ecosystem gross primary productivity (GPP). Solar-induced chlorophyll fluorescence (SIF) offers a good proxy for GPP since it marks the actual process of photosynthesis; while machine learning (ML) provides a robust approach to model the GPP-SIF relationship. Here, we trained the boosted regressing tree (BRT) and the Random Forest ML models with Greenhouse Gases Observing Satellite SIF data and in situ GPP observations from 49 eddy covariance towers. These trained ML GPP-SIF models were fed into the Energy Exascale Earth System Model (E3SM) Land Model (ELM) to generate ELM-simulated global SIF estimates, which were then benchmarked against satellite SIF observations with a surrogate modeling approach. Our results indicated good modeling performance of the ML-based GPP-SIF relationship. The ELM model when fed with the ML GPP-SIF models also can well predict the spatial-temporal variations in SIF. We also found high model accuracy for the surrogate modeling. Model parameter sensitivity analysis suggested that the fraction of leaf nitrogen in RuBisCO (flnr) is the most sensitive parameter to the SIF; other sensitive parameters include the Ball-Berry stomatal conductance slope (mbbopt) and the vcmax entropy (vcmaxse). The posterior uncertainty in simulated GPP was greatly reduced after benchmarking, and the model produced improved spatial patterns of mean GPP relative to FLUXCOM GPP. Our integrated approach provides a new avenue for improving land models and using remote-sensing SIF, which can be further improved in the future with more ground- and satellite-based observations.

54 ENVIRONMENTAL SCIENCES↗

Unveiling the principle descriptor for predicting the electron inelastic mean free path based on a machine learning framework

The TPP-2M formula is the most popular empirical formula for the estimation of the electron inelastic mean free paths (IMFPs) in solids from several simple material parameters. The TPP-2M formula, however, poorly describes several materials because it relies heavily on the traditional least-squares analysis. Herein, we propose a new framework based on machine learning to overcome the weakness. This framework allows a selection from an enormous number of combined terms (descriptors) to build a new formula that describes the electron IMFPs. The resulting framework not only provides higher average accuracy and stability but also reveals the physics meanings of several newly found descriptors. Using the identified principle descriptors, a complete physics picture of electron IMFPs is obtained, including both single and collective electron behaviors of inelastic scattering. Our findings suggest that machine learning is robust and efficient to predict the IMFP and has great potential in building a regression framework for data-driven problems. Furthermore, this method could be applicable to find empirical formula for given experimental data using a series of parameters given a priori, holds potential to find a deeper connection between experimental data and a priori parameters.

36 MATERIALS SCIENCE↗

Extracting the Galactic Center excess’ source-count distribution with neural nets

The two leading hypotheses for the Galactic Center excess (GCE) in the Fermi data are an unresolved population of faint millisecond pulsars (MSPs) and dark-matter (DM) annihilation. The dichotomy between these explanations is typically reflected by modeling them as two separate emission components. However, point sources (PSs) such as MSPs become statistically degenerate with smooth Poisson emission in the ultrafaint limit (formally where each source is expected to contribute much less than one photon on average), leading to an ambiguity that can render questions such as whether the emission is PS-like or Poissonian in nature ill defined. We present a conceptually new approach that describes the PS and Poisson emission in a unified manner and only afterwards derives constraints on the Poissonian component from the so obtained results. For the implementation of this approach, we leverage deep learning techniques, centered around a neural network-based method for histogram regression that expresses uncertainties in terms of quantiles. We demonstrate that our method is robust against a number of systematics that have plagued previous approaches, in particular DM/PS misattribution. In the Fermi data, we find a faint GCE described by a median source-count distribution (SCD) peaked at a flux of ~ 4 x 10 -11 counts cm -2 s -1 (corresponding to ~ 3-4 expected counts per PS), which would require N ~ $\mathscr{O}$(10 4 ) sources to explain the entire excess (median value N = 29,300 across the sky). Although faint, this SCD allows us to derive the constraint η ρ ≤ 66% for the Poissonian fraction of the GCE flux η ρ at 95% confidence, suggesting that a substantial amount of the GCE flux is due to PSs.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Intra-hour Solar Irradiance Forecast in Multiple Locations using Deep Transfer Learning

In recent years, solar power system installation imposes several challenges on the operations of local and regional power grids due to the inherent variability of ground-level solar irradiance. This work proposes a novel real-time solar forecast methodology for intra-hour solar irradiance based on deep transfer learning from ground-based sky imager for time horizons ranging from 5-15 min. There are three unique aspects of the proposed methodology: (1) a Deep Learning based algorithm development which is modeled as a classification approach rather than a traditional regression approach; (2) the use of the Transfer Learning technique to show generalization capability, robustness, and portability of baseline model in the newly deployed location where availability of enough data for training is typically scarce, and (3) redefinition of point-based irradiation forecast error estimation technique with a window-based one that is more intuitive and user-friendly. The system is developed using multiple years of irradiance and sky image recording in New Jersey and one-year data from Colorado, USA. The method is validated against ground telemetry from these two locations of diverse geographic and climatic conditions. Results show that the forecasting method proposed in this work is robust and highly accurate (8% MAPE error) for multiple locations deployment.

Deep Learning, Convolution Neural Networks, transf↗

Analysis of Present Day and Future OH and Methane Lifetime in the ACCMIP Simulations

Results from simulations performed for the Atmospheric Chemistry and Climate Modeling Intercomparison Project (ACCMIP) are analysed to examine how OH and methane lifetime may change from present day to the future, under different climate and emissions scenarios. Present day (2000) mean tropospheric chemical lifetime derived from the ACCMIP multi-model mean is 9.8+/-1.6 yr (9.3+/-0.9 yr when only including selected models), lower than a recent observationally-based estimate, but with a similar range to previous multi-model estimates. Future model projections are based on the four Representative Concentration Pathways (RCPs), and the results also exhibit a large range. Decreases in global methane lifetime of 4.5 +/- 9.1% are simulated for the scenario with lowest radiative forcing by 2100 (RCP 2.6), while increases of 8.5+/-10.4% are simulated for the scenario with highest radiative forcing (RCP 8.5). In this scenario, the key driver of the evolution of OH and methane lifetime is methane itself, since its concentration more than doubles by 2100 and it consumes much of the OH that exists in the troposphere. Stratospheric ozone recovery, which drives tropospheric OH decreases through photolysis modifications, also plays a partial role. In the other scenarios, where methane changes are less drastic, the interplay between various competing drivers leads to smaller and more diverse OH and methane lifetime responses, which are difficult to attribute. For all scenarios, regional OH changes are even more variable, with the most robust feature being the large decreases over the remote oceans in RCP8.5. Through a regression analysis, we suggest that differences in emissions of non-methane volatile organic compounds and in the simulation of photolysis rates may be the main factors causing the differences in simulated present day OH and methane lifetime. Diversity in predicted changes between present day and future OH was found to be associated more strongly with differences in modelled temperature and stratospheric ozone changes. Finally, through perturbation experiments we calculated an OH feedback factor (F) of 1.24 from present day conditions (1.50 from 2100 RCP8.5 conditions) and a climate feedback on methane lifetime of 0.33+-0.13 yr/K, on average. Models that did not include interactive stratospheric ozone effects on photolysis showed a stronger sensitivity to climate, as they did not account for negative effects of climate-driven stratospheric ozone recovery on tropospheric OH, which would have partly offset the overall OH/methane lifetime response to climate change.

atmospheric composition↗

Functional Data Analysis Applied to Modeling of Severe Acute Mucositis and Dysphagia Resulting From Head and Neck Radiation Therapy

Purpose: Current normal tissue complication probability modeling using logistic regression suffers from bias and high uncertainty in the presence of highly correlated radiation therapy (RT) dose data. This hinders robust estimates of dose-response associations and, hence, optimal normal tissue–sparing strategies from being elucidated. Using functional data analysis (FDA) to reduce the dimensionality of the dose data could overcome this limitation. Methods and Materials: FDA was applied to modeling of severe acute mucositis and dysphagia resulting from head and neck RT. Functional partial least squares regression (FPLS) and functional principal component analysis were used for dimensionality reduction of the dose-volume histogram data. The reduced dose data were input into functional logistic regression models (functional partial least squares–logistic regression [FPLS-LR] and functional principal component–logistic regression [FPC-LR]) along with clinical data. This approach was compared with penalized logistic regression (PLR) in terms of predictive performance and the significance of treatment covariate–response associations, assessed using bootstrapping. Results: The area under the receiver operating characteristic curve for the PLR, FPC-LR, and FPLS-LR models was 0.65, 0.69, and 0.67, respectively, for mucositis (internal validation) and 0.81, 0.83, and 0.83, respectively, for dysphagia (external validation). The calibration slopes/intercepts for the PLR, FPC-LR, and FPLS-LR models were 1.6/−0.67, 0.45/0.47, and 0.40/0.49, respectively, for mucositis (internal validation) and 2.5/−0.96, 0.79/−0.04, and 0.79/0.00, respectively, for dysphagia (external validation). The bootstrapped odds ratios indicated significant associations between RT dose and severe toxicity in the mucositis and dysphagia FDA models. Cisplatin was significantly associated with severe dysphagia in the FDA models. None of the covariates was significantly associated with severe toxicity in the PLR models. Dose levels greater than approximately 1.0 Gy/fraction were most strongly associated with severe acute mucositis and dysphagia in the FDA models. Conclusions: FPLS and functional principal component analysis marginally improved predictive performance compared with PLR and provided robust dose-response associations. FDA is recommended for use in normal tissue complication probability modeling.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Robust Time-Varying Parameter Identification for Composite Load Modeling

With the increasing integration of uncertain resources, e.g., renewables, electric vehicles, and demand responses, it is imperative to understand the characteristics of loads for power system analysis and control. Challenges of load modeling come from a variety of load components and time-varying compositions. In addition, the existence of outliers in measurements further complicates the problem. This paper proposes a robust time-varying parameter identification technique for composite ZIP and induction motor load models. A batch-model regression form, including time-varying parameters, is established based on state transition models and observation models with current observations and previous predictions to guarantee data redundancy. To deal with outliers, down-weighting coefficients of measurements are calculated with a projection statistics approach. Based on the batch-model regression and the down-weighting coefficients, parameter identification at each sample time is formulated as a weighted least squares optimization, which is solved by the Newton-Raphson approach with the previous estimated parameters as the initial iteration values. In addition, parameters' sensitivities to different outliers in measurements are analyzed. Results on the IEEE 57-bus system and IEEE 118-bus system show that the proposed algorithm can robustly identify time-varying parameters for the composite load models.

Composite load model↗

Refining T c Prediction in Hydrides via Symbolic‐Regression‐Enhanced Electron‐Localization‐Function‐Based Descriptors

Hydrogen‐based materials are able to possess extremely high superconducting critical temperatures, T c s , due to hydrogen's low atomic mass and strong electron–phonon interaction. Recently, a descriptor based on the Electron Localization Function (ELF) has enabled the rapid estimation of the T c of hydrogen‐containing compounds from electronic networking properties, but its applicability has been limited by the small size and homogeneity of the training dataset used. Herein, the model is re‐examined, compiling a publicly available combined dataset of 244 binary and ternary hydride superconductors. The analysis shows that though ELF‐based networking remains a valuable descriptor, its predictive power declines with increasing compositional complexity. However, by introducing the molecularity index, defined as the highest value of the ELF at which two hydrogen atoms connect, and applying symbolic regression, the accuracy of the predictions can be substantially enhanced. These results establish a more robust framework for assessing superconductivity in hydride materials, facilitating accelerated screening of novel candidates through integration with crystal structure prediction methods or high‐throughput searches.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Data augmentation for disruption prediction via robust surrogate models

The goal of this work is to generate large statistically representative data sets to train machine learning models for disruption prediction provided by data from few existing discharges. Such a comprehensive training database is important to achieve satisfying and reliable prediction results in artificial neural network classifiers. Here, we aim for a robust augmentation of the training database for multivariate time series data using Student t process regression. We apply Student t process regression in a state space formulation via Bayesian filtering to tackle challenges imposed by outliers and noise in the training data set and to reduce the computational complexity. Thus, the method can also be used if the time resolution is high. We use an uncorrelated model for each dimension and impose correlations afterwards via colouring transformations. We demonstrate the efficacy of our approach on plasma diagnostics data of three different disruption classes from the DIII-D tokamak. To evaluate if the distribution of the generated data is similar to the training data, we additionally perform statistical analyses using methods from time series analysis, descriptive statistics and classic machine learning clustering algorithms.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗