Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model selection”

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 613 records · Page 34

Best Practices for Wake Model and Optimization Algorithm Selection in Wind Farm Layout Optimization: Preprint

This paper presents the results of two case studies regarding the wind farm layout optimization problem. We asked a general audience to take part in the studies that we designed, and nine individuals participated. Case study 1 considered variations in optimization strategies for a given simple Gaussian wake model. Participants were provided with a wake model that outputs annual energy production (AEP) for an input set of wind turbine locations. Participants used an optimization method of their choosing to find an optimal wind farm layout. Case study 2 looked at trade-offs in performance resulting from variation in both physics model and optimization strategy. For case study 2, participants calculated AEP using a wake model of their choice while also implementing their chosen optimization method. Participants then used their wake model to calculate the AEP all other participant submitted turbine configurations produce for cross-comparison results. for optimized turbine locations were then cross-compared by recalculating the AEP using every other participant's wake model. Results for case study 1 show that the best optimal wind farm layouts in this study were achieved by participants who used gradient-based optimization methods. A front-runner emerged with the Sparse Nonlinear OPTimizer plus Wake Expansion Continuation (SNOPTplusWEC) optimization method, which consistently discovered a higher AEP for each scenario. Results for case study 2 show that for small wind farms with few turbines, turbine placement on the wind farm boundary is superior. Conclusions for case study 2 were drawn from participant cross-comparison of results.

17 WIND ENERGY↗

Utah FORGE: Earth Model Mesh Data for Selected Surfaces

This submission contains a number of data files with vertices of meshed/interpolated surfaces used in the Phase 2B earth model. Examples include land surface (based on 10-meter DEM), the granitoid-basin fill contact, several faults, and also interpolated temperature isosurfaces for 175 and 225 degrees C. All data are georeferenced to UTM, zone 12N, NAD 83, NAVD 88.

15 GEOTHERMAL ENERGY↗

Evaluating causal‐based feature selection for fuel property prediction models

Abstract In‐silico screening of novel biofuel molecules based on chemical and fuel properties is a critical first step in the biofuel evaluation process due to the significant volumes of samples required for experimental testing, the destructive nature of engine tests, and the costs associated with bench‐scale synthesis of novel fuels. Predictive models are limited by training sets of few existing measurements, often containing similar classes of molecules that represent just a subset of the potential molecular fuel space. Software tools can be used to generate every possible molecular descriptor for use as input features, but most of these features are largely irrelevant and training models on datasets with higher dimensionality than size tends to yield poor predictive performance. Feature selection has been shown to improve machine learning models, but correlation‐based feature selection fails to provide scientific insight into the underlying mechanisms that determine structure–property relationships. The implementation of causal discovery in feature selection could potentially inform the biofuel design process while also improving model prediction accuracy and robustness to new data. In this study, we investigate the benefits causal‐based feature selection might have on both model performance and identification of key molecular substructures. We found that causal‐based feature selection performed on par with alternative filtration methods, and that a structural causal model provides valuable scientific insights into the relationships between molecular substructures and fuel properties.

Nguyen, Bernard↗

Numerical characterization of support recovery in sparse regression with correlated design

Sparse regression is employed in diverse scientific settings as a feature selection method. A pervasive aspect of scientific data is the presence of correlations between predictive features. These correlations hamper both feature selection and estimation and jeopardize conclusions drawn from estimated models. On the other hand, theoretical results on sparsity-inducing regularized regression have largely addressed conditions for selection consistency via asymptotics, and disregard the problem of model selection, whereby regularization parameters are chosen. In this numerical study, we address these issues through exhaustive characterization of the performance of several regression estimators, coupled with a range of model selection strategies. These estimators and selection criteria were examined across correlated regression problems with varying degrees of signal to noise, distributions of non-zero model coefficients, and model sparsity. Our results reveal a fundamental tradeoff between false positive and false negative control in all regression estimators and model selection criteria examined. Additionally, we numerically explore a transition point modulated by the signal-to-noise ratio and spectral properties of the design covariance matrix at which the selection accuracy of all considered algorithms degrades. Overall, we find that SCAD coupled with BIC or empirical Bayes model selection performs the best feature selection across the regression problems considered.

97 MATHEMATICS AND COMPUTING↗

Selective Oxidation of Lignin Model Compounds

Lignin, the planet's most abundant renewable source of aromatic compounds, is difficult to degrade efficiently to welldefined aromatics. We developed a microwave-assisted catalytic Swern oxidation system using an easily prepared catalyst, MoO 2 Cl 2 (DMSO) 2 , and DMSO as the solvent and oxidant. It demonstrated high efficiency in transforming lignin model compounds containing the units and functional groups found in native lignins. The aromatic ring substituents strongly influenced the selectivity of β-ether phenolic dimer cleavage to generate sinapaldehyde and coniferaldehyde, monomers not usually produced by oxidative methods. Time-course studies on two key intermediates provided insight into the reaction pathway. Finally, owing to the broad scope of this oxidation system and the insight gleaned with regard to its mechanism, this strategy could be adapted and applied in a general sense to the production of useful aromatic chemicals from phenolics and lignin.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Multi-scale computational modeling of residual stress in selective laser melting with uncertainty quantification

Selective laser melting (SLM) is a powder-based additive manufacturing technique which creates parts by fusing together successive layers of powder with a laser. The quality of produced parts is highly dependent on the proper selection of processing parameters, requiring significant testing and experimentation to determine parameters for a given machine and material. Computational modeling could potentially be used to shorten this process by identifying parameters through simulation. However, simulating complete SLM builds is challenging due to the difference in scale between the size of the particles and laser used in the build and the size of the part produced. Often, continuum models are employed which approximate the powder as a continuous medium to avoid the need to model powder particles individually. While computationally expedient, continuum models require as inputs effective material properties for the powder which are often difficult to obtain experimentally. Building on previous works which have developed methods for estimating these effective properties along with their uncertainties through the use of detailed models, this work presents a part scale continuum model capable of predicting residual thermal stresses in an SLM build with uncertainty estimates. Model predictions are compared to experimental measurements from the literature.

42 ENGINEERING↗

Evaluation of hydrologic components of community land model 4 and bias identification

Runoff and soil moisture are two key components of the global hydrologic cycle that should be validated at local to global scales in Earth System Models (ESMs) used for climate projection. Here, we have evaluated the runoff and surface soil moisture output by the Community Climate System Model (CCSM) along with 8 other models from the Coupled Model Intercomparison Project (CMIP5) repository using satellite soil moisture observations and stream gauge corrected runoff products. A series of Community Land Model (CLM) runs forced by reanalysis and coupled model outputs was also performed to identify atmospheric drivers of biases and uncertainties in the CCSM. Results indicate that surface soil moisture simulations tend to be positively biased in high latitude areas by most selected CMIP5 models except CCSM, FGOALS, and BCC, which share similar land surface model code. With the exception of GISS, runoff simulations by all selected CMIP5 models were overestimated in mountain ranges and in most of the Arctic region. In general, positive biases in CCSM soil moisture and runoff due to precipitation input error were offset by negative biases induced by temperature input error. Excluding the impact from atmosphere modeling, the global mean of seasonal surface moisture oscillation was out of phase compared to observations in many years during 1985–2004. The CLM also underestimated runoff in the Amazon, central Africa, and south Asia, where soils all have high clay content. We hypothesize that lack of a macropore flow mechanism is partially responsible for this underestimation. However, runoff was overestimated in the areas covered by volcanic ash soils (i.e., Andisols), which might be associated with poor soil porosity representation in CLM. Finally, our results indicate that CCSM predictability of hydrology could be improved by addressing the compensating errors associated with precipitation and temperature and updating the CLM soil representation.

54 ENVIRONMENTAL SCIENCES↗

Minipig and beagle animal model genomes aid species selection in pharmaceutical discovery and development

Improving drug attrition remains a challenge in pharmaceutical discovery and development. A major cause of early attrition is the demonstration of safety signals which can negate any therapeutic index previously established. Safety attrition needs to be put in context of clinical translation (i.e. human relevance) and is negatively impacted by differences between animal models and human. In order to minimize such an impact, an earlier assessment of pharmacological target homology across animal model species will enhance understanding of the context of animal safety signals and aid species selection during later regulatory toxicology studies. Here we sequenced the genomes of the Sus scrofa Göttingen minipig and the Canis familiaris beagle, two widely used animal species in regulatory safety studies. Comparative analyses of these new genomes with other key model organisms, namely mouse, rat, cynomolgus macaque, rhesus macaque, two related breeds (S. scrofa Duroc and C. familiaris boxer) and human reveal considerable variation in gene content. Key genes in toxicology and metabolism studies, such as the UGT2 family, CYP2D6, and SLCO1A2, displayed unique duplication patterns. Comparisons of 317 known human drug targets revealed surprising variation such as species-specific positive selection, duplication and higher occurrences of pseudogenized targets in beagle (41 genes) relative to minipig (19 genes). These data will facilitate the more effective use of animals in biomedical research. - Highlights: • Genomes of the minipig and beagle dog, two species used in pharmaceutical studies. • First systematic comparative genome analysis of human and six experimental animals. • Key drug toxicology genes display unique duplication patterns across species. • Comparison of 317 drug targets show species-specific evolutionary patterns.

60 APPLIED LIFE SCIENCES↗

Developing an Evolutionary Baseline Model for Humans: Jointly Inferring Purifying Selection with Population History

Building evolutionarily appropriate baseline models for natural populations is not only important for answering fundamental questions in population genetics—including quantifying the relative contributions of adaptive versus nonadaptive processes—but also essential for identifying candidate loci experiencing relatively rare and episodic forms of selection (e.g., positive or balancing selection). Here, a baseline model was developed for a human population of West African ancestry, the Yoruba, comprising processes constantly operating on the genome (i.e., purifying and background selection, population size changes, recombination rate heterogeneity, and gene conversion). Specifically, to perform joint inference of selective effects with demography, an approximate Bayesian approach was employed that utilizes the decay of background selection effects around functional elements, taking into account genomic architecture. This approach inferred a recent 6-fold population growth together with a distribution of fitness effects that is skewed towards effectively neutral mutations. Importantly, these results further suggest that, although strong and/or frequent recurrent positive selection is inconsistent with observed data, weak to moderate positive selection is consistent but unidentifiable if rare.

59 BASIC BIOLOGICAL SCIENCES↗

Analyzing the impact of modeling choices and assumptions in compartmental epidemiological models

In this study, computational models have become increasingly used as part of modeling, predicting, and understanding how infectious diseases spread within large populations. These models can be broadly classified into differential equation-based models (EBM) and agent-based models (ABM). Both types of models are central in aiding public health officials design intervention strategies in case of large epidemic outbreaks. We examine these models in the context of illuminating their hidden assumptions and the impact these may have on the model outcomes. Very few ABM/EBMs are evaluated for their suitability to address a particular public health concern, and drawing relevant conclusions about their suitability requires reliable and relevant information regarding the different modeling strategies and associated assumptions. Hence, there is a need to determine how the different modeling strategies, choices of various parameters, and the resolution of information for EBMs and ABMs affect outcomes, including predictions of disease spread. In this study, we present a quantitative analysis of how the selection of model types (i.e., EBM vs. ABM), the underlying assumptions that are enforced by model types to model the disease propagation process, and the choice of time advance (continuous vs. discrete) affect the overall outcomes of modeling disease spread. Our study reveals that the magnitude and velocity of the simulated epidemic depends critically on the selection of modeling principles, various assumptions of disease process, and the choice of time advance.

60 APPLIED LIFE SCIENCES↗

Development and validation of an enhanced filtered drag model for simulating gas-solid fluidization of Geldart A particles in all flow regimes

Coarse-grid two-fluid simulation of gas-solid fluidized bed reactors based on the kinetic theory of granular flow exhibits a significant dependence on drag models, especially for Geldart A particles. Many drag models are available in the literature, which have been reported to work for different systems. This work focused on the evaluation of an enhanced filtered drag model along with other different drag models derived from different methods for three-dimensional two-fluid model simulations of gas-solid fluidized beds of Geldart A particles covering a broad range of fluidization regimes, including bubbling fluidization, turbulent fluidization, fast fluidization, and dilute phase transport regimes. Eight drag models were selected, which included five heterogeneous drag models and three homogeneous drag models. Comparison with the available experimental data demonstrates the need for modification of homogeneous drag models to account for the effect of mesoscale structures (i.e., bubbles and clusters). The enhanced filtered drag model and energy-minimization multi-scale (EMMS) drag models were found to achieve superior predictions in all fluidization regimes, while the other drag models were only capable of predicting certain fluidization regimes. Finally, the results of this work provide a guideline for choosing appropriate drag models for simulating Geldart A particles and suggestions on developing more reliable and general drag models applicable in all flow regimes.

42 ENGINEERING↗

Building thermal dynamics modeling with deep transfer learning using a large residential smart thermostat dataset

Understanding thermal dynamics and obtaining the computational model of residential buildings enable its scaled application in energy retrofits, control optimization and decarbonization. In this paper, we present a deep learning approach to model building thermal dynamics with smart thermostat data collected from residential buildings, with the goal to investigate model generalizability. In the first stage, we developed and compared different Deep Learning architectures including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) models and CNN-LSTM to predict indoor air temperature in a multi-step time horizon. In the second stage, we implemented a Transfer Learning (TL) process, which aims to improve the prediction performance on a new set of buildings (targets), exploiting the knowledge of related or similar buildings (sources). Different TL strategies and source model identification methods were investigated. The study showed that the CNN-LSTM performed the best among the architectures compared, with an average Mean Absolute Error (MAE) of 0.26 °C for one-hour-ahead (twelve 5-min future steps) predictions. Furthermore, the results showed that freezing the LSTM layer and fine-tuning the other layers of the CNN-LSTM achieved the best performance among four TL strategies, which further improved the performance with respect to a machine learning approach by 10%, and proving the effectiveness and generalizability of the proposed approach. A comparison of three different source model identification methods showed that randomly selecting source models constrained by similar building characteristics can provide good TL performance while retaining simplicity comparing with other quantitative source identification methods.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

SWAT model application for evaluating agricultural conservation practice effectiveness in reducing phosphorous loss from the Western Lake Erie Basin

Lake Erie is threatened by eutrophication and harmful algal blooms due to excess nutrient loading from agricultural sources. Agricultural conservation practices (ACPs) have been developed and implemented to reduce nutrient losses but estimating ACP effectiveness is challenging. The Soil and Water Assessment Tool (SWAT) has been used to investigate ACP effectiveness for water quality improvement. Many SWAT applications have been developed by different investigators to evaluate ACP effectiveness for reducing nutrient, particularly phosphorus (P), loading in the agriculturally-dominated Western Lake Erie Basin (WLEB). Our objective is to establish what has been achieved by past modeling research and make suggestions for future applications and improvements. We synthesized the findings of 28 SWAT modeling studies within the WLEB. Models generally performed satisfactorily against accepted criteria for streamflow and sediment, but performance for P loads, like soluble reactive P, was mostly “unsatisfactory”. The “unsatisfactory” performance maybe due to imperfections and idealizations in model formulations and/or parameterization. Thus, simulations of P transport and transformation processes need improvement. In addition, model parameter selection is the key part of model set-up. Most SWAT modeling studies used default values during initial set-up, then performed calibration and validation. It was found that the calibrated P related parameter values varied widely across different studies, even within the same watershed with some values unrealistic for the study areas. The phenomena of different combinations of model parameters producing similar outputs indicates equifinality. Equifinality in the baseline model may impact results when ACPs are incorporated. Furthermore, the unrealistic values used in ACP assessment undermine the credibility of ACP effectiveness. Future model applications should try to re-examine the calibrated P parameters and make sure they are realistic for the study area as well as reduce equifinality by constraining the model with characterization of watershed conditions, better understanding of hydrologic processes, and parameter values based on real-world observations. As a result, future model applications should focus on improving P transport and transformation processes, using measured watershed characteristics for parameterization, and improving reflections of climate change, which could result in more accurate assessments of ACP effectiveness to meet targeted goals.

54 ENVIRONMENTAL SCIENCES↗

Future sea level change under CMIP5 and CMIP6 scenarios from the Greenland and Antarctic ice sheets

Projections of the sea level contribution from the Greenland and Antarctic ice sheets rely on atmospheric and oceanic drivers obtained from climate models. The Earth System Models participating in the Coupled Model Intercomparison Project phase 6 (CMIP6) generally project greater future warming compared with the previous CMIP5 effort. Here we use four CMIP6 models and a selection of CMIP5 models to force multiple ice sheet models as part of the Ice Sheet Model Intercomparison Project for CMIP6 (ISMIP6). We find that the projected sea level contribution at 2100 from the ice sheet model ensemble under the CMIP6 scenarios falls within the CMIP5 range for the Antarctic ice sheet but is significantly increased for Greenland. Warmer atmosphere in CMIP6 models results in higher Greenland mass loss due to surface melt. For Antarctica, CMIP6 forcing is similar to CMIP5 and mass gain from increased snowfall counteracts increased loss due to ocean warming.

54 ENVIRONMENTAL SCIENCES↗

Machine learning approaches for influenza A virus risk assessment identifies predictive correlates using ferret model in vivo data

In vivo assessments of influenza A virus (IAV) pathogenicity and transmissibility in ferrets represent a crucial component of many pandemic risk assessment rubrics, but few systematic efforts to identify which data from in vivo experimentation are most useful for predicting pathogenesis and transmission outcomes have been conducted. To this aim, we aggregated viral and molecular data from 125 contemporary IAV (H1, H2, H3, H5, H7, and H9 subtypes) evaluated in ferrets under a consistent protocol. Three overarching predictive classification outcomes (lethality, morbidity, transmissibility) were constructed using machine learning (ML) techniques, employing datasets emphasizing virological and clinical parameters from inoculated ferrets, limited to viral sequence-based information, or combining both data types. Among 11 different ML algorithms tested and assessed, gradient boosting machines and random forest algorithms yielded the highest performance, with models for lethality and transmission consistently better performing than models predicting morbidity. Comparisons of feature selection among models was performed, and highest performing models were validated with results from external risk assessment studies. Our findings show that ML algorithms can be used to summarize complex in vivo experimental work into succinct summaries that inform and enhance risk assessment criteria for pandemic preparedness that take in vivo data into account.

59 BASIC BIOLOGICAL SCIENCES↗

Machine-learning-based dynamic-importance sampling for adaptive multiscale simulations

Multiscale simulations are a well-accepted way to bridge the length and time scales required for scientific studies with the solution accuracy achievable through available computational resources. Traditional approaches either solve a coarse model with selective refinement or coerce a detailed model into faster sampling, both of which have limitations. Here, we present a paradigm of adaptive, multiscale simulations that couple different scales using a dynamic-importance sampling approach. Our method uses machine learning to dynamically and exhaustively sample the phase space explored by a macro model using microscale simulations and enables an automatic feedback from the micro to the macro scale, leading to a self-healing multiscale simulation. As a result, our approach delivers macro length and time scales, but with the effective precision of the micro scale. Our approach is arbitrarily scalable as well as transferable to many different types of simulations. Overall, our method made possible a multiscale scientific campaign of unprecedented scale to understand the interactions of RAS proteins with a plasma membrane in the context of cancer research running over several days on Sierra, which is currently the second-most-powerful supercomputer in the world.

59 BASIC BIOLOGICAL SCIENCES↗

Optimization-Based Dynamic Voltage Support of Microgrids Using Energy Storage Systems

A microgrid network is characterized by a high R/X ratio, making the voltage more sensitive to active power changes compared to bulk power systems, where the voltage is regulated primarily by reactive power. Due to its sensitivity, voltage control approaches for microgrids should also consider the active power input coupling, making it very different from conventional power systems. Additionally, as the energy costs associated with active and reactive powers are different and the operational conditions of microgrids connected to active distribution systems vary over time, the ideal controller to provide voltage support must be flexible enough to handle these technical and operational constraints. This paper proposes a model predictive control approach to provide dynamic voltage support using energy storage systems. This approach uses a simplified predictive model of the system to solve the model predictive control problem. By proper selection of model predictive control weighting parameters, the quality of service provided can be adjusted to achieve the desired performance. A simulation study in MATLAB/Simulink validates the proposed approach for the Cordova, Alaska microgrid. Results show that the performance of the voltage support can be adjusted depending on the choice of weight and constraints of the controller.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Accurate, Fast and Scalable Kernel Ridge Regression on Parallel and Distributed Systems

Kernel Ridge Regression (KRR) is a fundamental method in machine learning. Given an n-by-d data matrix as input, a traditional implementation requires Θ(n2) memory to form an n-by-n kernel matrix and Θ(n3) flops to compute the final model. These time and storage costs prohibit KRR from scaling up to large datasets. For example, even on a relatively small dataset (a 520k-by-90 input requiring 357 MB), KRR requires 2 TB memory just to store the kernel matrix. Additionally, the reason is that n usually is much larger than d for real-world applications. On the other hand, weak scaling becomes a problem: if we keep d and n/p fixed as p grows (p is # machines), the memory needed grows as Θ(p) per processor and the flops as Θ(p2) per processor. In the perfect weak scaling situation, both the memory needed and the flops grow as Θ(1) per processor (i.e. memory and flops are constant). The traditional Distributed KRR implementation (DKRR) only achieved 0.32% weak scaling efficiency from 96 to 1536 processors. In this work, we propose two new methods to address these problems: the Balanced KRR (BKRR) and K-means KRR (KKRR). These methods consider alternative ways to partition the input dataset into p different parts, generating p different models, and then selecting the best model among them. Compared to a conventional implementation, KKRR2 (optimized version of KKRR) improves the weak scaling efficiency from 0.32% to 38% and achieves a 591x speedup for getting the same accuracy by using the same data and the same hardware (1536 processors). BKRR2 (optimized version of BKRR) achieves a higher accuracy than the current fastest method using less training time for a variety of datasets. For the applications requiring only approximate solutions, BKRR2 improves the weak scaling efficiency to 92% and achieves 3505x speedup (theoretical speedup: 4096x).

97 MATHEMATICS AND COMPUTING↗