Multiple Model Inference: Calibration and Selection with Multiple Models.
Abstract not provided.
SEARCH · Engineering Papers
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.
Abstract not provided.
ABSTRACT For topical, dermatological drug products, an in vitro option to determine bioequivalence (BE) between test and reference products is recommended. In particular, in vitro permeation test (IVPT) data analysis uses a reference‐scaled approach for two primary endpoints, cumulative penetration amount (AMT) and maximum flux ( J max ), which takes the within donor variability into consideration. In 2022, the Food and Drug Administration (FDA) published a draft IVPT guidance that includes statistical analysis methods for both balanced and unbalanced cases of IVPT study data. This work presents a comprehensive evaluation of various methodologies used to estimate critical parameters essential in assessing BE. Specifically, we investigate the performance of the FDA draft IVPT guidance approach alongside alternative empirical and model‐based methods utilizing mixed‐effects models. Our analyses include both simulated scenarios and real‐world studies. In simulated scenarios, empirical formulas consistently demonstrate robustness in approximating the true model, particularly in effectively addressing treatment–donor interactions. Conversely, the effectiveness of model‐based approaches heavily relies on precise model selection, which significantly influences their results. The research emphasizes the importance of accurate model selection in model‐based BE assessment methodologies. It sheds light on the advantages of empirical formulas, highlighting their reliability compared to model‐based approaches and offers valuable implications for BE assessments. Our findings underscore the significance of robust methodologies and provide essential insights to advance their understanding and application in the assessment of BE, employed in IVPT data analysis.
Model materials consisting of metal nanoparticles loaded onto oxide supports were synthesized, characterized, and investigated in a number of catalytic chemical reactions. By varying the size, shape, and composition of nanoparticle, as well as the material used to support the nanoparticles, it was found that small changes to the catalyst can have enormous changes to the reaction activity and selectivity. Investigation of these carefully synthesized catalysts via in situ characterization, and reaction studies, leads to a deeper understanding of the molecular level parameters that govern catalysis. Through study of the properties of the nanoparticles it was discovered that nanoparticle size and shape have a dominant role in the chemoselective catalysis of furfural over platinum nanoparticles. When vapor phase furfural and hydrogen gas were passed over Pt nanoparticles ranging in size from 1.5 to 7.1 nm, the catalytic selectivity was found to be dominated by the size of the nanoparticle. Large nanoparticles promoted hydrogenation of furfural to furfuryl alcohol, while smaller nanoparticles favored decarbonylation to furan. The same size specific selectivity was found in the hydrogenative reforming (the transformation of hydrocarbons to branched isomers) of C6 hydrocarbons, in which Pt nanoparticle size controls isomerization selectivity. Methylcyclopentane was found to be extremely size dependent at lower temperatures (553 K). It was found that smaller sized nanoparticles favored isomer formation, while larger sizes catalyzed the aromatization reaction more efficiently. n-hexane was found to be much less dependent on particle size, but still showed an increase in isomerization with small particles over larger sized Pt nanoparticles. The composition of PtxRh1-x bimetallic nanoparticles was also studied. These catalysts were characterized under hexane reforming conditions with Ambient Pressure X-ray Photoelectron Spectroscopy (AP-XPS), in order to find the actual surface atomic composition under real catalytic working conditions. By using AP-XPS and catalytic data in tandem, it was found that an optimum Rh loading occurred when the surface ensemble statistically favored one Rh atom surrounded by Pt atoms. By utilizing different oxide materials for catalytic supports the flow of charge can play a role in the reaction at the surface or interface in a phenomenon known as the strong metal-support interaction (SMSI). When Pt nanoparticles were loaded onto mesoporous supports made of Co3O4, NiO, MnO2, Fe2O3, and CeO2 it was found that their activity for carbon monoxide oxidation was greatly enhanced relative to the support alone or Pt loaded onto inert mesoporous silica. This finding demonstrates that the interface of the metallic Pt nanoparticle and the oxide 2 support is able to produce turnovers that are orders of magnitude higher than the two materials separately. When the same type of experiments were investigated with n-hexane as the reactant and macroporous Al2O3, TiO2, Nb2O5, Ta2O5, and ZrO2 were utilized as supports, it was found that the reaction selectivity was greatly altered depending on the catalytic support material. TiO2, Nb2O5, and Ta2O5 (all of which are strong Lewis acids) were found to be much more selective for isomer production than the standard SiO2 mesoporous silica supported Pt nanoparticle catalyst. Finally, an acidified mesoporous silica material was utilized as the support. This material was synthesized by using AlCl3 to modify the surface of mesoporous silica. This support was found to have no activity for hexane isomerization alone. However, when Pt nanoparticles were supported on the material, the activity and isomer selectivity in hexane reforming was increased several orders of magnitude as compared to the same nanoparticles supported on unmodified mesoporous silica. This dissertation builds on the existing knowledge of known concepts in catalysis science such as structure sensitive reactions, the metal-support interaction, and acid-base chemistry. The results show how small changes in the active sites of a catalyst can create large changes in the catalytic chemistry. This research demonstrates how careful material control, characterization and reaction study can help to elucidate the molecular level components necessary to design efficient catalysts.
When conducting model averaging for assessing groundwater conceptual model uncertainty, the averaging weights are often evaluated using model selection criteria such as AIC, AICc, BIC, and KIC (Akaike Information Criterion, Corrected Akaike Information Criterion, Bayesian Information Criterion, and Kashyap Information Criterion, respectively). However, this method often leads to an unrealistic situation in which the best model receives overwhelmingly large averaging weight (close to 100%), which cannot be justified by available data and knowledge. It was found in this study that this problem was caused by using the covariance matrix, CE, of measurement errors for estimating the negative log likelihood function common to all the model selection criteria. This problem can be resolved by using the covariance matrix, Cek, of total errors (including model errors and measurement errors) to account for the correlation between the total errors. An iterative two-stage method was developed in the context of maximum likelihood inverse modeling to iteratively infer the unknown Cek from the residuals during model calibration. The inferred Cek was then used in the evaluation of model selection criteria and model averaging weights. While this method was limited to serial data using time series techniques in this study, it can be extended to spatial data using geostatistical techniques. The method was first evaluated in a synthetic study and then applied to an experimental study, in which alternative surface complexation models were developed to simulate column experiments of uranium reactive transport. It was found that the total errors of the alternative models were temporally correlated due to the model errors. The iterative two-stage method using Cek resolved the problem that the best model receives 100% model averaging weight, and the resulting model averaging weights were supported by the calibration results and physical understanding of the alternative models. Using Cek obtained from the iterative two-stage method also improved predictive performance of the individual models and model averaging in both synthetic and experimental studies.
Abstract Objective . To propose a mathematical model for applying ionization detail (ID), the detailed spatial distribution of ionization along a particle track, to proton and ion beam radiotherapy treatment planning (RTP). Approach . Our model provides for selection of preferred ID parameters ( I p ) for RTP, that associate closest to biological effects. Cluster dose is proposed to bridge the large gap between nanoscopic I p and macroscopic RTP. Selection of I p is demonstrated using published cell survival measurements for protons through argon, comparing results for nineteen I p : N k , k = 2, 3, …, 10, the number of ionizations in clusters of k or more per particle, and F k , k = 1, 2, …, 10, the number of clusters of k or more per particle. We then describe application of the model to ID-based RTP and propose a path to clinical translation. Main results . The preferred I p were N 4 and F 5 for aerobic cells, N 5 and F 7 for hypoxic cells. Significant differences were found in cell survival for beams having the same LET or the preferred N k . Conversely, there was no significant difference for F 5 for aerobic cells and F 7 for hypoxic cells, regardless of ion beam atomic number or energy. Further, cells irradiated with the same cluster dose for these I p had the same cell survival. Based on these preliminary results and other compelling results in nanodosimetry, it is reasonable to assert that I p exist that are more closely associated with biological effects than current LET-based approaches and microdosimetric RBE-based models used in particle RTP. However, more biological variables such as cell line and cycle phase, as well as ion beam pulse structure and rate still need investigation. Significance . Our model provides a practical means to select preferred I p from radiobiological data, and to convert I p to the macroscopic cluster dose for particle RTP.
We consider the possibility that the dark energy is made up of two or more independent components, each having a different equation of state. We fit the model with supernova and gamma-ray burst data from recent observations, and use the Markov Chain Monte Carlo technique to estimate the allowed parameter regions. We also use various model selection criteria to compare the two-component model with the {lambda}CDM, one-component dark energy model with static or variable w (XCDM) and other multicomponent models. We find that the two-component models can give a reasonably good fit to the current data. For some data sets, and depending somewhat on the model selection criteria, the two-component model can give a better fit to the data than XCDM with static w and XCDM with variable w parametrized by w=w{sub 0}+w{sub a}z/(1+z)
We study the problem of selecting between different mechanisms of supersymmetry breaking in the minimal supersymmetric standard model using current data. We evaluate the Bayesian evidence of four supersymmetry breaking scenarios: mSUGRA, mGMSB, mAMSB, and moduli mediation. The results show a strong dependence on the dark matter assumption. Using the inferred cosmological relic density as an upper bound, minimal anomaly mediation is at least moderately favored over the CMSSM. Our fits also indicate that evidence for a positive sign of the {mu} parameter is moderate at best. We present constraints on the anomaly and gauge mediated parameter spaces and some previously unexplored aspects of the dark matter phenomenology of the moduli mediation scenario. We use sparticle searches, indirect observables and dark matter observables in the global fit and quantify robustness with respect to prior choice. We quantify how much information is contained within each constraint.
Sim-SEQ is an international initiative on model comparison for geologic carbon sequestration, with an objective to understand and, if possible, quantify model uncertainties. Model comparison efforts in Sim-SEQ are at present focusing on one specific field test site, hereafter referred to as the Sim-SEQ Study site (or S-3 site). Within Sim-SEQ, different modeling teams are developing conceptual models of CO2 injection at the S-3 site. In this paper, we select five flow models of the S-3 site and provide a qualitative comparison of their attributes and predictions. These models are based on five different simulators or modeling approaches: TOUGH2/EOS7C, STOMP-CO2e, MoReS, TOUGH2-MP/ECO2N, and VESA. In addition to model-to-model comparison, we perform a limited model-to-data comparison, and illustrate how model choices impact model predictions. We conclude the paper by making recommendations for model refinement that are likely to result in less uncertainty in model predictions.
NASA's Global Ecosystem Dynamics Investigation (GEDI) is collecting spaceborne full waveform lidar data with a primary science goal of producing accurate estimates of forest aboveground biomass density (AGBD). This paper presents the development of the models used to create GEDI's footprint-level (~25 m) AGBD (GEDI04_A) product, including a description of the datasets used and the procedure for final model selection. The data used to fit our models are from a compilation of globally distributed spatially and temporally coincident field and airborne lidar datasets, whereby we simulated GEDI-like waveforms from airborne lidar to build a calibration database. We used this database to expand the geographic extent of past waveform lidar studies, and divided the globe into four broad strata by Plant Functional Type (PFT) and six geographic regions. GEDI's waveform-to-biomass models take the form of parametric Ordinary Least Squares (OLS) models with simulated Relative Height (RH) metrics as predictor variables. From an exhaustive set of candidate models, we selected the best input predictor variables, and data transformations for each geographic stratum in the GEDI domain to produce a set of comprehensive predictive footprint-level models. We found that model selection frequently favored combinations of RH metrics at the 98th, 90th, 50th, and 10th height above ground-level percentiles (RH98, RH90, RH50, and RH10, respectively), but that inclusion of lower RH metrics (e.g. RH10) did not markedly improve model performance. Second, forced inclusion of RH98 in all models was important and did not degrade model performance, and the best performing models were parsimonious, typically having only 1-3 predictors. Third, stratification by geographic domain (PFT, geographic region) improved model performance in comparison to global models without stratification. Fourth, for the vast majority of strata, the best performing models were fit using square root transformation of field AGBD and/or height metrics. There was considerable variability in model performance across geographic strata, and areas with sparse training data and/or high AGBD values had the poorest performance. These models are used to produce global predictions of AGBD, but will be improved in the future as more and better training data become available.
A transient three dimensional model for describing the molten pool dynamics and the response of oxidation film evolution in the selective laser melting of aluminum-based material is proposed. The physical difference in both sides of the scan track, powder-solid transformation and temperature dependent physical properties are taken into account. It shows that the heat energy tends to accumulate in the powder material rather than in the as-fabricated part, leading to the formation of the asymmetrical patterns of the temperature contour and the attendant larger dimensions of the molten pool in the powder phase. As a higher volumetric energy density is applied (≥1300 J/mm{sup 3}), a severe evaporation is produced with the upward direction of velocity vector in the irradiated powder region while a restricted operating temperature is obtained in the as-fabricated part. The velocity vector continuously changes from upward direction to the downward one as the scan speed increases from 100 mm/s to 300 mm/s, promoting the generation of the debris of the oxidation films and the resultant homogeneous distribution state in the matrix. For the applied hatch spacing of 50 μm, a restricted remelting phenomenon of the as-fabricated part is produced with the upward direction of the convection flow, significantly reducing the turbulence of the thermal-capillary convection on the breaking of the oxidation films, and therefore, the connected oxidation films through the neighboring layers are typically formed. The morphology and distribution of the oxidation are experimentally acquired, which are in a good agreement with the results predicted by simulation.
Abstract The purpose of this study is to use the PERSIANN–Climate Data Record (PERSIANN-CDR) dataset to evaluate the ability of 32 CMIP5 models in capturing the behavior of daily extreme precipitation estimates globally. The daily long-term historical global PERSIANN-CDR allows for a global investigation of eight precipitation indices that is unattainable with other datasets. Quantitative comparisons against CPC daily gauge; GPCP One-Degree Daily (GPCP1DD); and TRMM 3B42, version 7 (3B42V7), datasets show the credibility of PERSIANN-CDR to be used as the reference data for global evaluation of CMIP5 models. This work uniquely defines different study regions by partitioning global land areas into 25 groups based on continent and climate zone type. Results show that model performance in warm temperate and equatorial regions in capturing daily extreme precipitation behavior is largely mixed in terms of index RMSE and correlation, suggesting that these regions may benefit from weighted model averaging schemes or model selection as opposed to simple model averaging. The three driest climate regions (snow, polar, and arid) exhibit high correlations and low RMSE values when compared against PERSIANN-CDR estimates, with the exceptions of the cold regions showing an inability to capture the 95th and 99th percentile annual total precipitation characteristics. A comprehensive assessment of each model’s performance in each continent–climate zone defined group is provided as a guide for both model developers to target regions and processes that are not yet fully captured in certain climate types, and for climate model output users to be able to select the models and/or the study areas that may best fit their applications of interest.
Nuclear power is a relatively carbon-free energy source that has the capacity to be utilized today in an effort to stem the tides of global warming. The growing demand for nuclear energy, however, could put significant strain on our uranium ore resources, and the mining activities utilized to extract that ore can leave behind long-term environmental damage. A potential solution to enhance the supply of uranium fuel is to recover uranium from seawater using amidoximated adsorbent fibers. This technology has been studied for decades but is currently plagued by the material’s relatively poor selectivity of uranium over its main competitor vanadium. In this work, we investigate the binding schemes between uranium, vanadium, and the amidoxime functional groups on the adsorbent surface. Using quantum chemical methods, binding strengths are approximated for a set of complexation reactions between uranium and vanadium with amidoxime functionalities. Those approximations are then coupled with a comprehensive aqueous adsorption model developed in this work to simulate the adsorption of uranium and vanadium under laboratory conditions. Experimental adsorption studies with uranium and vanadium over a wide pH range are performed, and the data collected are compared against simulation results to validate the model. It was found that coupling ab initio calculations with process level adsorption modeling provides accurate predictions of the adsorption capacity and selectivity of the sorbent materials. Furthermore, this work demonstrates that this multiscale modeling paradigm could be utilized to aid in the selection of superior ligands or ligand compositions for the selective capture of metal ions. Furthermore, this first-principles integrated modeling approach opens the door to the in silico design of next-generation adsorbents with potentially superior efficiency and selectivity for uranium over vanadium in seawater.
This report demonstrates the capabilities of the net-zero microgrid (NZM) Xendee platform for modeling an SR module with electricity, heat extraction and thermal storage in microgrids configurations. The model effectively captures the most important technical and economic considerations for SR technology specific analysis: cost and operational characteristics of SR technology and financial costs and incentives. The model can analyze multiple scenarios to establish metrics for cost-competitive and zero-carbon microgrids connected to the grid or completely isolated. The model is fully integrated within the Xendee platform for modeling and analysis of clean energy microgrids with storage and generation from renewable energy sources. The model captures the capabilities, constraints, and nuances of SR by incorporating parameters related to plant economics, design efficiency and performance, plant operation and component and fuel lifespan. The cost and operational parameters modeled in the SR module are specific to the technology selected for integration in the microgrid. Cost parameters recognize advanced nuclear technology for modular production and installation based on economies of scale from factory manufacture and related commissioning, and cost reduction through technology maturation—first-of-a-kind (FOAK) and nth-of-a-Kind (NOAK). The cost parameters include installation, operations and maintenance (O&M), fuel refueling cycle, and reactor life. Installation cost reflects economies of scale due to unit sizing at scale and colocation. O&M economies of scale for both fixed- and variable-cost fuel life-cycle costs are incurred at every refueling interval, with separate front- and back-end fuel costs, as well as waste-handling and disposition costs. This report investigates key characteristics of different SR technologies suitable for microgrid applications, including design principles, sizing, coolant properties, temperature ratings, fuel structures, and life-cycle considerations. This also includes fuel technologies applicable to these SR systems, alongside strategies for nuclear-waste and spent-fuel management and approaches to address safety, security, and proliferation challenges. Four primary groups of SR technologies are examined: water-cooled, liquid-metal-cooled, high-temperature gas-cooled, and molten-salt-cooled systems. In this report, an initial guideline for technology selection is established, aligning the characteristics of the technologies with the requirements of microgrids. The selection of technology in a microgrid is influenced by various factors, including financial capacity, location and accessibility, demand type and characteristics, reliability and resilience requirements, area constraints, and the lifespan of the microgrid. The types of electrical and non-electrical applications within the microgrid also play a significant role in technology selection. The characteristics of SRs, such as their smaller size, modularity, transportability, long refueling interval, improved safety features, ability to operate in autonomous or semi-autonomous mode, and provision of high-grade heat, are particularly appealing for microgrids. Furthermore, a list of considerations for implementing SRs in microgrids is outlined. The SR model is created to be continuously improved with the acquisition of actual data on investment and operational costs, experience with supply chains, production at scale, and field deployments. In the near term, performance data on applications in microgrids will become available from lessons learned from laboratory tests, such as those planned for the Microreactor Applications Research Validation and Evaluation Project (MARVEL), led by Idaho National Laboratory (INL). The SR model incorporates scenario data and known SR design specifications, enabling technoeconomic analysis for SR deployment in microgrids. It specifically considers the distinctive attributes of SRs as generators in technoeconomic studies. SRs can be modeled and analyzed with generation from renewable-energy sources, energy storage, and flexible loads over a range of functionality and applications. This offers a comprehensive tool for feasibility studies, scenario development, and sensitivity analysis for “what-if” consideration of any range of assumptions about SRs in microgrids and other aggregations of distributed-energy resources, including virtual power plants.
Purpose: Post-irradiation cerebral necrosis (PICN) is a severe late effect that can Result from brain cancers treatment using radiation therapy. The purpose of this study was to compare the treatment plans and predicted risk of PICN after volumetric modulated arc therapy (VMAT) to the risk after passively scattered proton therapy (PSPT) and intensity modulated proton therapy (IMPT) in a cohort of pediatric patients. Methods: Thirteen pediatric patients with varying age and sex were selected for this study. A clinical treatment volume (CTV) was constructed for 8 glioma patients and 5 ependymoma patients. Prescribed dose was 54 Gy over 30 fractions to the planning volume. Dosimetric endpoints were compared between VMAT and proton plans. The normal tissue complication probability (NTCP) following VMAT and proton therapy planning was also calculated using PICN as the biological endpoint. Sensitivity tests were performed to determine if predicted risk of PICN was sensitive to positional errors, proton range errors and selection of risk models. Results: Both PSPT and IMPT plans resulted in a significant increase in the maximum dose and reduction in the total brain volume irradiated to low doses compared with the VMAT plans. The average ratios of NTCP between PSPT and VMAT were 0.56 and 0.38 for glioma and ependymoma patients respectively and the average ratios of NTCP between IMPT and VMAT were 0.67 and 0.68 for glioma and ependymoma plans respectively. Sensitivity test revealed that predicted ratios of risk were insensitive to range and positional errors but varied with risk model selection. Conclusion: Both PSPT and IMPT plans resulted in a decrease in the predictive risk of necrosis for the pediatric plans studied in this work. Sensitivity analysis upheld the qualitative findings of the risk models used in this study, however more accurate models that take into account dose and volume are needed.
Abstract Induced earthquakes often accompany fluid injection, and the seismic hazard they pose threatens various underground engineering projects. Models to monitor and control induced seismic hazard with traffic light systems should be probabilistic, forward‐looking, and updated as new data arrive. In this study, we propose an Induced Seismicity Test Bench to test and rank such models; this test bench can be used for model development, model selection, and ensemble model building. We apply the test bench to data from the Basel 2006 and Soultz‐sous‐Forêts 2004 geothermal stimulation projects, and we assess forecasts from two models: Shapiro and Smoothed Seismicity (SaSS) and Hydraulics and Seismics (HySei). These models incorporate a different mix of physics‐based elements and stochastic representation of the induced sequences. Our results show that neither model is fully superior to the other. Generally, HySei forecasts the seismicity rate better after shut‐in but is only mediocre at forecasting the spatial distribution. On the other hand, SaSS forecasts the spatial distribution better and gives better seismicity rate estimates before shut‐in. The shut‐in phase is a difficult moment for both models in both reservoirs: the models tend to underpredict the seismicity rate around, and shortly after, shut‐in.
Microstructure characterization enables the development of structure-processing-property relationships critical to several research areas within the broad field of materials science, from alloy design to the assessment of corrosion resistance, and failure analysis. Conventional approaches to material characterization have relied on either qualitative inference by the human ex-pert or software applications that can extract high-level features from images, such as boundary segmentation, average grain diameter, etc. Such approaches rely heavily on subject matter expert user intervention and knowledge of what phases or more generally, what microstructural features, are of interest. The recent surge in the adoption of machine learning techniques to address problems in materials engineering has brought with it an increased interest and application of Image Driven Machine Learning (IDML) approaches. In this work, we review the applications of IDML to the field of materials characterization. A canonical hierarchy of stages is defined, which when put sequentially together completes an IDML study: problem definition, dataset building, model selection and training, model evaluation, and integration with existing instrumentation or simulation workflow. The studies reviewed in this work are analyzed from the perspective of each of these stages. Such a review permits agranular assessment of the field, for example the impact of IDML on materials characterization at the nanoscale, the size of a typical dataset required to train a semantic segmentation model on electron microscopy images, ubiquitousness of transfer learning in the domain, etc. Finally, we discuss the importance of interpretability and explainability in the field of IDML for materials characterization, and provide an overview of two emerging techniques in the field: semantic segmentation and generative adversarial networks.
Four-dimensional (4D) scanning transmission electron microscopy is one of the most rapidly growing modes of electron microscopy imaging. The advent of fast pixelated cameras and the associated data infrastructure have greatly accelerated this process. Yet conversion of the 4D datasets into physically meaningful structure images in real space remains an open issue. In this work, we demonstrate that it is possible to systematically create filters that will affect the apparent resolution or even qualitative features of the real-space structure image, reconstructing artificially generated patterns. As initial efforts, we explore statistical model selection algorithms, aiming for robustness and reliability of estimated filters. This statistical model selection analysis demonstrates the need for regularization and cross validation of inversion methods to robustly recover structure from high-dimensional diffraction datasets.
Abstract not provided.