AI Surrogate Model for Distributed Computing Workloads
Explore the source record for details and available documents.
97 MATHEMATICS AND COMPUTING↗
SEARCH · Engineering Papers
Results for “probabilistic analysis”
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.
A probabilistic inverse prediction method for predicting plutonium processing conditions
In the past decade, nuclear chemists and physicists have been conducting studies to investigate the signatures associated with the production of special nuclear material (SNM). In particular, these studies aim to determine how various processing parameters impact the physical, chemical, and morphological properties of the resulting special nuclear material. By better understanding how these properties relate to the processing parameters, scientists can better contribute to nuclear forensics investigations by quantifying their results and ultimately shortening the forensic timeline. This paper aims to statistically analyze and quantify the relationships that exist between the processing conditions used in these experiments and the various properties of the nuclear end-product by invoking inverse methods. In particular, these methods make use of Bayesian Adaptive Spline Surface models in conjunction with Bayesian model calibration techniques to probabilistically determine processing conditions as an inverse function of morphological characteristics. Not only does the model presented in this paper allow for providing point estimates of a sample of special nuclear material, but it also incorporates uncertainty into these predictions. This model proves sufficient for predicting processing conditions within a standard deviation of the observed processing conditions, on average, provides a solid foundation for future work in predicting processing conditions of particles of special nuclear material using only their observed morphological characteristics, and is generalizable to the field of chemometrics for applicability across different materials.
11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
Uniformly Ordered Binary Decision Algorithm for Benchmark Experiment Correlations in Whisper Validation
When performing a validation exercise for determining the upper subcritical limit of a nuclear criticality safety application, an analyst should select and perform a statistical analysis on a population of benchmark experiments that are neutronically similar to the application. The size of this population should be sufficiently large such that the statistical analysis has a high degree of confidence that the bias plus bias uncertainty (calculational margin) has been accurately quantified. A complication arises because many benchmark experiments share common components, leading to correlations in their measured effective multiplication factors. Correlations between benchmark experiments within the population reduces its predictive power. This motivates the need for methods that consider benchmark experiment correlations and ensure adequate statistical significance of results. The Whisper code is a statistical analysis pack- age that incorporates nuclear data sensitivity coefficients from MCNP to assess benchmark experiment similarity and then performs an extreme-value analysis to estimate the bias plus bias uncertainty. The original methodology in Whisper does not consider the effect of benchmark experiment correlations when making this estimation, and this summary proposes the uniformly ordered binary decision algorithm to address this shortcoming. The original methodology in Whisper computes similarity coefficients ck for an application compared to all benchmark experiments in its library and develops weighting factors for a selected population proportional to the ck values. The methodology can be interpreted as statistically emulating a validation exercise for a particular application where the weighting factors may be viewed as the likelihood that an analyst would include a particular benchmark experiment within the population. The effective sample size of the population is the expected or mean number of benchmark experiments in the population. The uniformly ordered binary decision algorithm identifies clusters of correlated benchmark experiments within the population and then computes adjusted weighting factors based on the magnitude of the correlation coefficients within the cluster to compute a reduced effective sample size accounting for the lower information content because of correlations. Benchmark experiments within the cluster are ordered randomly with equal probability and probabilistic decisions are made as to whether a benchmark. experiment within the cluster should treated as redundant with a previous one; if two redundant benchmark experiments are included, then the conservative worst case bias plus bias uncertainty is used and the pair is counted as a single benchmark experiment in the population. Results are provided for HEU solutions in a research version of the Whisper software using benchmark experiment correlations provided by DICE, the Database for the International Criticality Safety Benchmark Evaluation Project (ICSBEP). These show that there can be a significant increase in the bias plus bias uncertainty because the effective sample size is reduced, and therefore the algorithm, needing to meet sample size requirements, expands the benchmark experiment population by accepting less similar benchmark experiments that would have otherwise not been included.
73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
Enhanced Component Performance Study: Emergency Diesel Generators 1998–2018
This report presents an enhanced performance evaluation of emergency diesel generators (EDGs) at U.S. commercial nuclear power plants. This report evaluates component performance over time using (1) Institute of Nuclear Power Operations (INPO) Consolidated Events Database (ICES) data from 1998 through 2018 and (2) maintenance unavailability (UA) performance data from Mitigating Systems Performance Index (MSPI) Basis Document data from 2002 through 2018. The objective is to show estimates of current failure probabilities and rates related to EDGs, trend these data on an annual basis, determine if the current data are consistent with the probability distributions currently recommended for use in NRC probabilistic risk assessments, show how the reliability data differ for different EDG manufacturers and for EDGs with different ratings; and summarize the subcomponents, causes, detection methods, and recovery associated with each EDG failure mode. Engineering analyses were performed with respect to time period and failure mode without regard to the actual number of EDGs at each plant. The factors analyzed are: sub-component, failure cause, detection method, recovery, manufacturer, and EDG rating. A statistically significant increasing trend was identified in the frequency of FTLR demands for emergency power system (EPS) and high pressure core spray (HPCS) EDGs and a statistically significant decreasing trend was identified in the frequency of run > 1H hours for EPS and HPCS EDGs.
99 GENERAL AND MISCELLANEOUS↗
Efficient Contingency Analysis in Power Systems via Network Trigger Nodes
Modeling failure dynamics within a power system is a complex and challenging process due to multiple inter-dependencies and convoluted inter-domain relationships. Subject matter experts (SMEs) are interested in understanding these failure dynamics for reducing the impact from future disasters (i.e., losses or failures of power system components, such as transmission lines). Contingency analysis (CA) tools enable such ’what-if’ scenario analyses to evaluate the impacts on the power system. Analyzing all possible contingencies among N system components can be computationally expensive. An important step for performing CA is identifying a set of k ‘trigger’ components, which when failed initially can significantly impact the overall system by causing multiple failures. Currently SMEs focus on identifying these trigger components by running expensive simulations on all possible subsets, which quickly becomes infeasible. Hence finding a relevant set of trigger components (contingencies) rapidly to enable efficient and useful CA is crucial.In a collaboration between computer scientists and power system experts, we propose an efficient method for performing CA by exploiting network inter-dependencies in power system components. First, we construct a network with multiple electric grid infrastructure components and dependencies as connections among them. We reformulate the problem of finding a set of trigger components as a problem of identifying critical nodes in the network, which can cascade power failures through connected nodes and cause significant damage to the network. To guide the practical CA tools, we develop a network-based model with a probabilistic edge-weights setup using intricate domain rules. Then we conduct an empirical study on real power system data in the US for both regional and national levels. Firstly, we use power system datasets for the US to create a national-scale domain-driven model. Secondly, we demonstrate that network-based model outperforms the outputs from a real CA tool and show on average 25 × improved selection of contingencies, thereby showcasing practical benefits to the power experts.
Tabassum, Anika↗
Randomized Algorithms for Scientific Computing (RASC)
Randomized algorithms have propelled advances in artificial intelligence (AI) and represent a foundational research area in advancing AI for Science. Future advancements in DOE Office of Science priority areas such as climate science, astrophysics, fusion, advanced materials, combustion, and quantum computing all require randomized algorithms for surmounting challenges of complexity, robustness, and scalability. Advances in data collection and numerical simulation have changed the dynamics of scientific research and motivate the need for randomized algorithms. For instance, advances in imaging technologies such as X-ray ptychography, electron microscopy, electron energy loss spectroscopy, or adaptive optics lattice light-sheet microscopy collect hyperspectral imaging and scattering data in terabytes, at breakneck speed enabled by state-of-the-art detectors. The data collection is exceptionally fast compared with its analysis. Likewise, advances in high-performance architectures have made exascale computing a reality and changed the economies of scientific computing in the process. Floating-point operations that create data are essentially free in comparison with data movement. Thus far, most approaches have focused on creating faster hardware. Ironically, this faster hardware has exacerbated the problem by making data still easier to create. Under such an onslaught, scientists often resort to heuristic deterministic sampling schemes (e.g., low-precision arithmetic, sampling every nth element) and sacrifice potentially valuable accuracy. Dramatically better results can be achieved via randomized algorithms, reducing the data size as much as or more than naive deterministic subsampling can achieve, while retaining the high accuracy of computing on the full data set. By randomized algorithms we mean those algorithms that employ some form of randomness in internal algorithmic decisions to accelerate time to solution, increase scalability, or improve reliability. Examples include matrix sketching for solving large-scale least-squares problems (see Figure 1) and stochastic gradient descent for training machine learning models. We are not recommending heuristic methods but rather randomized algorithms that have certificates of correctness and probabilistic guarantees of optimality and near-optimality. Such approaches can be useful beyond acceleration, for example, in understanding how to avoid measure zero worst-case scenarios that plague methods such as QR matrix factorization.
97 MATHEMATICS AND COMPUTING↗
A Hybrid AI/ML and Computational Mechanics Based Approach for Time-Series State and Fatigue Life Estimation of Nuclear Reactor Components
Environmental fatigue modeling is a complex problem due to multiple failure modes and their intermixing. The failure modes are function of various underlying causes in addition to the corrosive effect of reactor coolant environment. Some of the major causes are time-dependence of material associated with cyclic loading, load sequence effect associated with random/variable amplitude loading, effect of strain amplitude and rates, effect of varying temperature (along both temporal and spatial directions) and the effect of mean strain and stress. The nonlinear intermixing of failure modes associated with above mentioned causing parameters makes the environmental fatigue modeling is a challenging task. Because of this challenge, fatigue is traditionally being modeled based on experimental data. However, test based empirical approach often requires hundreds of fatigue tests to model the above-mentioned intermixing failure causes even for a single material system. The problem is further exaggerated for reactor component made from multi-material systems such as made from both carbon and stainless-steel base metals and their similar and dissimilar metal welds. With the difficulty of conducting hundreds of fatigue tests to capture the above-mentioned intermixing failure causes, fatigue modeling approaches often depends on empirical models based on limited available test data such as available through ASME code and NUREG 6909. However, these limited test-data-based models may not be enough to accurately predict the life of reactor components. Accurate prediction of life of reactor component would become a necessity, particularly when the license of the reactors to be extended for long-term-operation (LTO) that is for well beyond its original design life of 40 years. The requirement of extending the license of reactor under LTO requires hundreds of fatigue tests to be conducted to understand the mechanism associated with the above-mentioned interdependent failure causes. However, conducting large number of fatigue tests is not a feasibility due to the cost involved. To address this issues Argonne National Laboratory (ANL) with the sponsorship of DOE Light Water Reactor Sustainability (LWRS) program trying to develop a hybrid predictive modeling approach. This is based on limited experiment-data, Artificial-intelligence (AI) – Machine-Learning (ML) - Deep-Learning (DL) based techniques and Multiphysics-computational-mechanics based modeling tools. The hybrid approach not-only can improve the accuracy of the existing stress analysis and fatigue modeling approach but also can reduce the over-dependency on test-based approach. Towards this goal following are some of the major contributions based on ANL’s FY-20 environmental fatigue modeling activities: 1) A cyclic plasticity material model database for 82/182 dissimilar metal weld, which can be readily shared with US nuclear industry and regulatory agency on request. 2) A well validated analytical modeling methodology to perform cycle-by-cycle stress prediction under both constant amplitude fatigue loading and variable amplitude fatigue loading (with load-sequence effect). 3) An AI/ML/DL based methodology to predict unmeasurable cyclic strain based on other available sensor signals. This type of approach can be used for estimating strain in real reactor components from other sensor readings. 4) An AI/ML based approach to improve the US capability on environmental fatigue testing. This is by improving ANL’s existing environmental fatigue testing capacity to conduct ASME required strain-controlled tests (by controlling strain amplitudes and its rate), while not measuring the strain (due to the difficulty of placing an extensometer in a narrow autoclave in a PWR-water-test system). 5) A simulation and experiment based probabilistic modeling methodology for time-series fatigue state and life estimation of reactor metal such as dissimilar metal weld.
22 GENERAL STUDIES OF NUCLEAR REACTORS↗
Dynamic Probabilistic Safety Assessment Studies for Advanced Reactor Using RAVEN
Probabilistic Safety Assessment (PSA) is used extensively to evaluate the risks associated with complex engineering systems like Nuclear Power Plants (NPPs). Current PSA models are based on the Event-Tree/Fault-Tree (ET/FT) methodology. ET and FT models are static and are based on Boolean logic approaches. In the past, concerns have been raised in the literature regarding the capability of the traditional static modelling approaches to adequately account for the impact of process, hardware, software, firmware and human interactions on the stochastic system behaviour. To overcome the limitations of the traditional approach to PSA, several dynamic PSA methodologies have been proposed. One of the dynamic PSA methodologies used for dynamic evaluations is Dynamic Event Tree (DET) framework which can be used to assess the impact of the parameter variability and scenario dynamics on the PSA model for the initiating event. The DET framework couples the stochastic model (number of component/trains that start on demand, operator action timing, etc.) with a Thermal-Hydraulic (TH) model of the plant. This paper explores the use of DET along with a case study on advanced reactor. The initiating event selected for the study was Class IV power supply failure event. The TH analysis considering uncertainty in various parameters was performed using RELAP5 and Reactor Analysis and Virtual control ENvironment (RAVEN) tool. Based on the uncertainty analysis, it is concluded that the peak clad temperatures (PCT) are within the limits in all the code runs implying a high-degree of safety margin. However, variation in time to reach the PCT was observed among the code runs and the mean time to reach the PCT was found to be around 8590sec (approximately 2.4 hours). Hence, sufficient time margin is available for human intervention and the operator might have a relatively stress-free state during such an accident scenario. Due to the static nature of the traditional PSA models, the safety margin available was lesser, whereas, with the help of dynamic PSA models, one can demonstrate that the actual available safety margin is more in the present case study and is valuable input from the design point of view.
99 GENERAL AND MISCELLANEOUS↗
Causes and evolution of winter polynyas north of Greenland
During the 42-year period (1979–2020) of satellite measurements, four major winter (December–March) polynyas have been observed north of Greenland: one in December 1986 and three in the last decade, i.e., February of 2011, 2017, and 2018. The 2018 polynya was unparalleled in its magnitude and duration compared to the three previous events. Given the apparent recent increase in the occurrence of these extreme events, this study aims to examine their evolution and causality, in terms of forced versus natural variability. The limited weather station and remotely sensed sea ice data are analyzed combining with output from the fully coupled Regional Arctic System Model (RASM), including one hindcast and two ensemble simulations. We found that neither the accompanying anomalous warm surface air intrusion nor the ocean below had an impact (i.e., no significant ice melting) on the evolution of the observed winter open-water episodes in the region. Instead, the extreme atmospheric wind forcing resulted in greater sea ice deformation and transport offshore, accounting for the majority of sea ice loss in all four polynyas. Our analysis suggests that strong southerly winds (i.e., northward wind with speeds greater than 10 m s –1 ) blowing persistently over the study region for at least 2 d or more were required over the study region to mechanically redistribute some of the thickest Arctic sea ice out of the region and thus to create open-water areas (i.e., a latent heat polynya). To assess the role of internal variability versus external forcing of such events, we carried out and examined results from the two RASM ensembles dynamically downscaled with output from the Community Earth System Model (CESM) Decadal Prediction Large Ensemble (DPLE) simulations. Out of 100 winters in each of the two ensembles (initialized 30 years apart: one in December 1985 and another in December 2015), 17 and 16 winter polynyas were produced north of Greenland, respectively. The frequency of polynya occurrence had no apparent sensitivity to the initial sea ice thickness in the study area pointing to internal variability of atmospheric forcing as a dominant cause of winter polynyas north of Greenland. We assert that dynamical downscaling using a high-resolution regional climate model offers a robust tool for process-level examination in space and time, synthesis with limited observations, and probabilistic forecasts of Arctic events, such as the ones being investigated here and elsewhere.
54 ENVIRONMENTAL SCIENCES↗
Modeling Framework for Bulk Electric Grid Impacts from HEMP E1 and E3 Effects (Tasks 3.1 Final Report)
This report presents a framework to evaluate the impact of a high-altitude electromagnetic pulse (HEMP) event on a bulk electric power grid. This report limits itself to modeling the impact of EMP E1 and E3 components. The co-simulation of E1 and E3 is presented in detail, and the focus of the paper is on the framework rather than actual results. This approach is highly conservative as E1 and E3 are not maximized with the same event characteristics and may only slightly overlap. The actual results shown in this report are based on a synthetic grid with synthetic data and a limited exemplary EMP model. The framework presented can be leveraged and used to analyze the impact of other threat scenarios, both manmade and natural disasters. This report d escribes a Monte-Carlo based methodology to probabilistically quantify the transient response of the power grid to a HEMP event. The approach uses multiple fundamental steps to characterize the system response to HEMP events, focused on the E1 and E3 components of the event. 1) Obtain component failure data related to HEMP events testing of components and creating component failure models. Use the component failure model to create component failure conditional probability density function (PDF) that is a function of the HEMP induced terminal voltage. 2) Model HEMP scenarios and calculate the E1 coupled voltage profiles seen by all system components. Model the same HEMP scenarios and calculate the transformer reactive power consumption profiles due to E3. 3) Sample each component failure PDF to determine which grid components will fail, due to the E1 voltage spike, for each scenario. 4) Perform dynamic simulations that incorporate the predicted component failures from E1 and reactive power consumption at each transformer affected by E3. These simulations allow for secondary transients to affect the relays/protection remaining in service which can lead to cascading outages. 5) Identify the locations and amount of load lost for each scenario through grid dynamic simulation. This can be an indication of the immediate grid impacts from a HEMP event. In addition, perform more detailed analysis to determine critical nodes and system trends. 6) To help realize the longer-term impacts, a security constrained alternating current optimal power flow (ACOPF) is run to maximize critical load served. This report describes a modeling framework to assess the systemic grid impacts due to a HEMP event. This stochastic simulation framework generates a large amount of data for each Monte Carlo replication, including HEMP location and characteristics, relay and component failures, E3 GIC profiles, cascading dynamics including voltage and frequency over time, and final system state. This data can then be analyzed to identify trends, e.g., unique system behavior modes or critical components whose failure is more likely to cause serious systemic effects. The proposed analysis process is demonstrated on a representative system. In order to draw realistic conclusions of the impact of a HEMP event on the grid, a significant amount of work remains with respect to modeling the impact on various grid components.
24 POWER TRANSMISSION AND DISTRIBUTION↗
Essential barrier height and a probabilistic approach in characterizing potential landscape
In this work we propose a probabilistic approach to investigate the shape of landscapes of multi-dimensional potential functions. Under a suitable coupling scheme, two copies of the overdamped Langevin dynamics associated with the potential function are coupled, and the coupling times are collected. Assuming a set of intuitive yet technically challenging conditions on the coupling scheme, it is shown that the tail distributions of the coupling times exhibit qualitatively different dependencies on the noise magnitude for single-well versus multi-well potential functions. More specifically, for convex single-well potentials, the negative tail exponent of the coupling time distribution is uniformly bounded away from zero by the convexity parameter and is independent of the noise magnitude. In contrast, for multi-well potentials, the negative tail exponent decreases exponentially as the noise vanishes, with the decay rate governed by the essential barrier height, a quantity introduced in this paper to characterize the non-convex nature of the potential function. Numerical investigations are conducted for a variety of examples, including the Rosenbrock function, interacting particle systems, and loss functions arising in artificial neural networks. These examples not only illustrate the theoretical results in various contexts but also provide crucial numerical validation of the conjectured assumptions, which are essential to the theoretical analysis yet lie beyond the reach of standard technical tools.
97 MATHEMATICS AND COMPUTING↗
flat10MIP: an emissions-driven experiment to diagnose the climate response to positive, zero and negative CO2 emissions
Abstract. The proportionality between global mean temperature and cumulative emissions of CO2 predicted in Earth system models (ESMs) is the foundation of carbon budgeting frameworks. Deviations from this behavior could impact estimates of required net-zero timings and negative emissions requirements to meet the Paris Agreement climate targets. However, existing ESM diagnostic experiments do not allow for direct estimation of these deviations as a function of defined emissions pathways. Here, we perform a set of climate model diagnostic experiments for the assessment of transient climate response to cumulative CO2 emissions (TCRE), the Zero Emissions Commitment (ZEC), and climate reversibility metrics in an emissions-driven framework. The emissions-driven experiments provide consistent independent variables simplifying simulation, analysis and interpretation, with emissions rates more comparable to recent levels than existing protocols using model-specific compatible emissions from the CMIP DECK 1pctCO2 experiment, where emissions rates tend to increase during the experiment, such that at the time of CO2 doubling in year 70, emissions are much greater than present-day values. A base experiment, “esm-flat10”, has constant emissions of CO2 of 10 GtC per year (near-present-day values), and initial results show that the TCRE estimated in this experiment is about 0.1 K less than that obtained using 1pctCO2. A subset of ESMs exhibit land carbon sinks that saturate during this experiment. A branch experiment, esm-flat10-zec, illustrates that both positive and negative ZEC effects are less pronounced under esm-flat10 than under 1pctCO2 – the magnitude of ZEC50 in ESMs is, on average, reduced by 30 % compared with 1pctCO2 branch experiments. A final experiment, esm-flat10-cdr, assesses climate reversibility under negative emissions, where we find that peak warming may occur before or after net zero and that the asymmetry in temperature at a given level of cumulative emissions between the positive and negative emissions phases is well described by ZEC in most models. Further, we find that existing probabilistic simple climate model (SCM) ensembles tend to overestimate temperature reversibility compared with ESMs, highlighting the need for additional constraints. We propose a set of climate diagnostic indicators to quantify various aspects of climate reversibility. These experiments were suggested as potential candidates in CMIP7 and have since been adopted as “fast track” simulations.
Sanderson, Benjamin M↗
Flexible Plant Operation and Generation Technical Program Plan for FY2023
This report presents the Technical Program Plan for Fiscal Years 2023-2027 (FY2023 to FY2027) for the U.S. Department of Energy (DOE) Light Water Reactor Sustainability Program—Flexible Plant Operation and Generation Research Pathway. The objective of this pathway is to carry out the research needed to help nuclear power plants diversify revenue generation for the life of these plants. The purpose of these research and development activities is two fold: (1) to reduce the technical and economic risks of implementing FPOG applications and (2) to provide guidance on relevant safety and environmental operating license reviews, amendments, and renewals. This pathway provides a clear understanding of the benefits of nuclear energy beyond electricity markets. A detailed description of the research and development activities that have been completed and that are planned for FY2023—FY2027 are presented in this report. These activities include completing the development of analysis tools to perform technical and economic assessments of realistic market opportunities for producing secondary energy products near nuclear power plants. They also include developing and demonstrating engineering systems and control concepts to dispatch thermal and electrical power to an industrial user. Additionally, this plan includes developing guidance for addressing potential regulatory and licensing requirements. In addition, the formation, purpose, and activities of a group referred to as the Hydrogen Regulatory Research and Review Group is discussed. An overview is also provided on the potential benefits of the Infrastructure Investment and Jobs Act (IIJA) Bill that will support the commencement of Regional Clean Hydrogen Hubs, and the Inflation Reduction Act (IRA) that provides compelling production tax credits for nuclear electricity and clean hydrogen using nuclear energy.
08 HYDROGEN↗
A data-driven method for modelling dissipation rates in stratified turbulence
We present a deep probabilistic convolutional neural network (PCNN) model for predicting local values of small-scale mixing properties in stratified turbulent flows, namely the dissipation rates of turbulent kinetic energy and density variance, $\varepsilon$ and $\chi$ . Inputs to the PCNN are vertical columns of velocity and density gradients, motivated by data typically available from microstructure profilers in the ocean. The architecture is designed to enable the model to capture several characteristic features of stratified turbulence, in particular the dependence of small-scale isotropy on the buoyancy Reynolds number $Re_b:=\varepsilon /(\nu N^2)$ , where $\nu$ is the kinematic viscosity and $N$ is the background buoyancy frequency, the correlation between suitably locally averaged density gradients and turbulence intensity and the importance of capturing the tails of the probability distribution functions of values of dissipation. Empirically modified versions of commonly used isotropic models for $\varepsilon$ and $\chi$ that depend only on vertical derivatives of density and velocity are proposed based on the asymptotic regimes $Re_b\ll 1$ and $Re_b\gg 1$ , and serve as an instructive benchmark for comparison with the data-driven approach. When trained and tested on a simulation of stratified decaying turbulence which accesses a range of turbulent regimes (associated with differing values of $Re_b$ ), the PCNN outperforms assumptions of isotropy significantly as $Re_b$ decreases, and additionally demonstrates improvements over the fitted empirical models. A differential sensitivity analysis of the PCNN facilitates a comparison with the theoretical models and provides a physical interpretation of the features enabling it to make improved predictions.
42 ENGINEERING↗
Unsupervised probabilistic models for sequential Electronic Health Records
We develop an unsupervised probabilistic model for heterogeneous Electronic Health Record (EHR) data. Utilizing a mixture model formulation, our approach directly models sequences of arbitrary length, such as medications and laboratory results. This allows for subgrouping and incorporation of the dynamics underlying heterogeneous data types. The model consists of a layered set of latent variables that encode underlying structure in the data. These variables represent subject subgroups at the top layer, and unobserved states for sequences in the second layer. We train this model on episodic data from subjects receiving medical care in the Kaiser Permanente Northern California integrated healthcare delivery system. The resulting properties of the trained model generate novel insight from these complex and multifaceted data. In addition, we show how the model can be used to analyze sequences that contribute to assessment of mortality likelihood.
59 BASIC BIOLOGICAL SCIENCES↗
MOOSE ProbML: Parallelizable Probabilistic Machine Learning and Uncertainty Quantification Capabilities
The Multiphysics Object Oriented Simulation Environment (MOOSE) is a widely used open- source finite element software for performing multiphysics multiscale simulations in a massively parallel fashion. Recently, the computational team at Idaho National Laboratory (INL) has implemented Probabilistic Machine Learning (ProbML) capabilities in MOOSE—in a parallelized fashion—and enable active learning with large-scale computational models for tasks such as surrogate model development, scale bridging, forward/inverse uncertainty quantification (UQ), Bayesian optimization, etc. This presentation summarizes these developments in MOOSE along with demonstrations on several real applications relevant to nuclear energy. At the fundamental level, samplers like Monte Carlo/Latin Hypercube, variance reduction, parallelized Markov Chain Monte Carlo (MCMC) support uncertainty propagation in both forward and inverse settings. These samplers can be integrated with the Gaussian processes (GP) suite in MOOSE, which offer several variants like scalar GPs, multi-output GPs, and deep GPs, to enable active learning. These GPs can be tuned using gradient-based optimization methods like Adam and its variants or gradient-free methods like the elliptical slice sampler (a variant of MCMC adept under Gaussian settings) for more complex covariance kernels or likelihoods whose gradient computations can be cumbersome. A variety of batch acquisition functions permit parallelized evaluation of the computational model and support different learning objectives with high efficiency like Bayesian inference, global surrogate development, optimization, etc. Furthermore, libtorch integration supports training, evaluation, and re-training of neural networks and other complex machine learning models in active learning settings. The impacts of these developments are shown on several real applications: (1) nuclear fuel inverse UQ and model inadequacy assessment using the Kennedy O’Hagan framework; (2) uncertainty aware surrogate modeling for additive manufacturing to predict field quantities; (3) nuclear reactor rare events analysis; and (4) complex fluid flow prediction using a global surrogate with quantified prediction uncertainty. Finally, the outlook of MOOSE ProbML is discussed for both outer-loop and inner-loop computations in the broad view to accelerate fuels and materials qualification, address gaps in knowledge and data, and assess new reactor/fuel systems.
11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
Deep Direct-Use Geothermal Feasibility Study for Hawthorne NV
The Hawthorne Nevada, deep direct-use geothermal study is a two-year effort funded by the U.S. Department of Energy to determine the techno-economic feasibility of implementing a large-scale, direct-use facility for the Hawthorne Army Depot (HAD) and the public facilities of the city of Hawthorne and Mineral County. The approach links a production side analysis (PSA) and a demand side analysis (DSA) into a whole-system analysis (WSA) to provide an integrated assessment of the resource and the probability of delivering economically viable direct-use energy to Hawthorne. Hawthorne, Nevada is in the western part of the Basin and Range province and has been the focus of geothermal investigations for over 40 years. Over the last 15 years, several studies completed by the U.S. Navy Geothermal Program Office (GPO) in conjunction with industry professionals quantified the existence of several low temperature geothermal prospects, the most promising of which is called Prospect A. The promise of Prospect A is based on drilling and flow testing that produced ~100 °C water at flow rates of up to 31 l/s (500 gallons per minute). Measured productivity indexes range from 40-85 l/s/MPa, suggesting a warm and productive heat source. Despite the promise of the resource, uncertainties in its spatial extent and long-term sustainability mean that techno-economic analyses must include probabilities of the sustainability of the resource under different operating scenarios. Here, the PSA is conducted by integrating a wide range of disparate data to estimate lognormal P90, P50, and P10 resource capacities. These capacities are used as input to a thermal-hydrologic (T-H) model to estimate thermal drawdown for each capacity estimate for several different DSA scenarios. Using a systems-based approach, the WSA links the dynamic T-H simulations of the PSA/DSA combinations with the techno-economic model GEOPHIRES to account for both the temporal dynamics and uncertainties in the system to produce probabilistic distributions of several performance metrics including the levelized cost of heat (LCOH) and the return on investment (ROI). This report is the final delivery for the project and documents the study's activities and results.
15 GEOTHERMAL ENERGY↗
Application of Point Precipitation Frequency Estimates to Watersheds
This report documents work sponsored by the U.S. Nuclear Regulatory Commission (NRC) at the Oak Ridge National Laboratory (ORNL) as part of the RES project, “Application of Point Precipitation Frequency Estimates to Watersheds.” This project was implemented as part of the Probabilistic Flood Hazard Assessment (PFHA) Research Program. The objective of the PFHA Research Program is to develop tools and guidance on the use of PFHA methods to risk-inform NRC’s licensing of new facilities as well as licensing and oversight of currently operating facilities as they relate to flooding hazards. Many nuclear power plants (NPPs) are located on or near rivers so riverine flooding hazards need to be considered in their design and operation. Probabilistic riverine flood models are important tools for realistic assessment of flooding risks. However, these models require areal estimates of the depth, duration, and frequency of rainfall distributed over the watershed, which are not often available. Point precipitation frequency estimates are more widely available. For example, the National Oceanic and Atmospheric Administration (NOAA) has published NOAA Atlas 14, which provides point precipitation frequency estimates for 5-minute through 60-day durations at average recurrence intervals of 1-year through 1,000-year. The research documented in this report addresses areal reduction factors (ARFs), which can be used to convert the widely available point precipitation frequency estimates, to estimates of areal precipitation frequency over a watershed. The most widely used ARF source is Technical Paper 29 (TP-29) published by the then U.S. Weather Bureau in 1958. However, both the methods and the underlying precipitation data used to produce TP-29 are seriously out of date. For example, due to the small gauge network available at the time of TP-29’s compilation, ARF estimates developed are only for watersheds smaller than about 400 square miles. Due to the relatively short record lengths of precipitation data available, frequency considerations could not be accurately determined. Other factors such as regional climate and seasonality were not addressed. Several newer methods have been published since TP-29 was developed and both the type and quantity of precipitation data have increased significantly, along with computational resources and analytical tools such as geographic information systems. This report reviewed and assessed the available precipitation products and methods for conducting ARF analysis. The work applied up-to-date precipitation data products and analysis methods with a novel watershed-based approach to investigate how ARF estimates vary across different methods, data sources, geographical locations, return periods, and seasons. The overall findings reported here regarding basic ARF trends are in line with other recent studies showing that ARFs decrease with increasing area, increase with increasing duration, and decrease with increasing return period. This study found significant differences among the available ARF methods. This work also found a strong geographical variability across different US hydrologic regions, suggesting that the ARF are specific to regional climate patterns and geographical characteristics and should not be applied arbitrarily to other locations. The results also reveal the importance of data record length, especially for high return level ARFs. The work reported in NUREG/CR-7271 will assist NRC staff in assessing different classes of ARF methods in conjunction with available rainfall data sets. It will also support the development of guidance for application of point precipitation data in PFHAs. It should be noted that the ARF values presented in this report for any location or region were developed for the purposes of comparing methods and investigating the factors that influence ARFs. They should not be considered official and should not be used in leu of a site-specific analysis.
54 ENVIRONMENTAL SCIENCES↗