Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “bayesian optimization”

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 955 records · Page 53

On the Uncertainty in Single Molecule Fluorescent Lifetime and Energy Emission Measurements

Time-correlated single photon counting has recently been combined with mode-locked picosecond pulsed excitation to measure the fluorescent lifetimes and energy emissions of single molecules in a flow stream. Maximum likelihood (ML) and least squares methods agree and are optimal when the number of detected photons is large, however, in single molecule fluorescence experiments the number of detected photons can be less than 20, 67 percent of those can be noise, and the detection time is restricted to 10 nanoseconds. Under the assumption that the photon signal and background noise are two independent inhomogeneous Poisson processes, we derive the exact joint arrival time probability density of the photons collected in a single counting experiment performed in the presence of background noise. The model obviates the need to bin experimental data for analysis, and makes it possible to analyze formally the effect of background noise on the photon detection experiment using both ML or Bayesian methods. For both methods we derive the joint and marginal probability densities of the fluorescent lifetime and fluorescent emission. The ML and Bayesian methods are compared in an analysis of simulated single molecule fluorescence experiments of Rhodamine 110 using different combinations of expected background noise and expected fluorescence emission. While both the ML or Bayesian procedures perform well for analyzing fluorescence emissions, the Bayesian methods provide more realistic measures of uncertainty in the fluorescent lifetimes. The Bayesian methods would be especially useful for measuring uncertainty in fluorescent lifetime estimates in current single molecule flow stream experiments where the expected fluorescence emission is low. Both the ML and Bayesian algorithms can be automated for applications in molecular biology.

Brown, Emery N.↗

Multiple objective optimization in reliability demonstration test

Reliability demonstration tests are usually performed in product design or validation processes to demonstrate whether a product meets specified requirements on reliability. For binomial demonstration tests, the zero-failure test has been most commonly used due to its simplicity and use of minimum sample size to achieve an acceptable consumer’s risk level. However, this test can often result in unacceptably high risk for producers as well as a low probability of passing the test even when the product has good reliability. This paper explicitly explores the interrelationship between multiple objectives that are commonly of interest when planning a demonstration test and proposes structured decision-making procedures using a Pareto front approach for selecting an optimal test plan based on simultaneously balancing multiple criteria. Different strategies are suggested for scenarios with different user priorities and graphical tools are developed to help quantify the trade-offs between choices and to facilitate informed decision making. As a result, potential impacts of some subjective user inputs on the final decision are studied to offer insights and useful guidance for general applications.

97 MATHEMATICS AND COMPUTING↗

Bayesian, Multifidelity Operator Learning for Complex Engineering Systems–A Position Paper

Abstract Deep learning has significantly improved the state-of-the-art in computer vision and natural language processing, and holds great potential to design effective tools for predicting and simulating complex engineering systems. In particular, scientific machine learning seeks to apply the power of deep learning to scientific and engineering tasks, with operator learning (OL) emerging as a particularly effective tool. OL can approximate nonlinear operators arising in complex engineering systems, making it useful for simulating, designing, and controlling those systems. In this position paper, we provide a comprehensive overview of OL, including its potential applications to complex engineering domains. We cover three variations of OL approaches: deterministic OL for modeling nonautonomous systems, OL with uncertainty quantification (UQ) capabilities, and multifidelity OL. For each variation, we discuss drawbacks and potential applications to engineering, in addition to providing a detailed explanation. We also highlight how multifidelity OL approaches with UQ capabilities can be used to design, optimize, and control engineering systems. Finally, we outline some potential challenges for OL within the engineering domain.

Computer Science↗

A Physical Model to Estimate Snowfall over Land using AMSU-B Observations

In this study, we present an improved physical model to retrieve snowfall rate over land using brightness temperature observations from the National Oceanic and Atmospheric Administration's (NOAA) Advanced Microwave Sounder Unit-B (AMSU-B) at 89 GHz, 150 GHz, 183.3 +/- 1 GHz, 183.3 +/- 3 GHz, and 183.3 +/- 7 GHz. The retrieval model is applied to the New England blizzard of March 5, 2001 which deposited about 75 cm of snow over much of Vermont, New Hampshire, and northern New York. In this improved physical model, prior retrieval assumptions about snowflake shape, particle size distributions, environmental conditions, and optimization methodology have been updated. Here, single scattering parameters for snow particles are calculated with the Discrete-Dipole Approximation (DDA) method instead of assuming spherical shapes. Five different snow particle models (hexagonal columns, hexagonal plates, and three different kinds of aggregates) are considered. Snow particle size distributions are assumed to vary with air temperature and to follow aircraft measurements described by previous studies. Brightness temperatures at AMSU-B frequencies for the New England blizzard are calculated using these DDA calculated single scattering parameters and particle size distributions. The vertical profiles of pressure, temperature, relative humidity and hydrometeors are provided by MM5 model simulations. These profiles are treated as the a priori data base in the Bayesian retrieval algorithm. In algorithm applications to the blizzard data, calculated brightness temperatures associated with selected database profiles agree with AMSU-B observations to within about +/- 5 K at all five frequencies. Retrieved snowfall rates compare favorably with the near-concurrent National Weather Service (NWS) radar reflectivity measurements. The relationships between the NWS radar measured reflectivities Z(sub e) and retrieved snowfall rate R for a given snow particle model are derived by a histogram matching technique. All of these Z(sub e)-R relationships fall in the range of previously established Z(sub e)-R relationships for snowfall. This suggests that the current physical model developed in this study can reliably estimate the snowfall rate over land using the AMSU-B measured brightness temperatures.

Kim, Min-Jeong↗

Reconstructing the Sky Location of Gravitational-Wave Detected Compact Binary Systems: Methodology for Testing and Comparison

The problem of reconstructing the sky position of compact binary coalescences detected via gravitational waves is a central one for future observations with the ground-based network of gravitational-wave laser interferometers, such as Advanced LIGO and Advanced Virgo. Different techniques for sky localization have been independently developed. They can be divided in two broad categories: fully coherent Bayesian techniques, which are high latency and aimed at in-depth studies of all the parameters of a source, including sky position, and "triangulation-based" techniques, which exploit the data products from the search stage of the analysis to provide an almost real-time approximation of the posterior probability density function of the sky location of a detection candidate. These techniques have previously been applied to data collected during the last science runs of gravitational-wave detectors operating in the so-called initial configuration. Here, we develop and analyze methods for assessing the self consistency of parameter estimation methods and carrying out fair comparisons between different algorithms, addressing issues of efficiency and optimality. These methods are general, and can be applied to parameter estimation problems other than sky localization. We apply these methods to two existing sky localization techniques representing the two above-mentioned categories, using a set of simulated inspiralonly signals from compact binary systems with a total mass of equal to or less than 20M solar mass and nonspinning components. We compare the relative advantages and costs of the two techniques and show that sky location uncertainties are on average a factor approx. equals 20 smaller for fully coherent techniques than for the specific variant of the triangulation-based technique used during the last science runs, at the expense of a factor approx. equals 1000 longer processing time.

technique used↗

Uncertainty quantification of a deep learning fuel property prediction model

Deep learning models are being widely used in the field of combustion. Given the black-box nature of typical neural network based models, uncertainty quantification (UQ) is critical to ensure the reliability of predictions as well as the training datasets, and for a principled quantification of noise and its various sources. Deep learning surrogate models for predicting properties of chemical compounds and mixtures have been recently shown to be promising for enabling data-driven fuel design and optimization, with the ultimate goal of improving efficiency and lowering emissions from combustion engines. In this study, UQ is performed for a multi-task deep learning model that simultaneously predicts the research octane number (RON), Motor Octane Number (MON), and Yield Sooting Index (YSI) of pure components and multicomponent blends. The deep learning model is comprised of three smaller networks: Extractor 1, Extractor 2, and Predictor, and a mixing operator. The molecular fingerprints of individual components are encoded via Extractor 1 and Extractor 2, the mixing operator generates fingerprints for mixtures/blends based on linear mixing operation, and the predictor maps the fingerprint to the target properties. Two different classes of UQ methods, Monte Carlo ensemble methods and Bayesian neural networks (BNNs), are employed for quantifying the epistemic uncertainty. Combinations of Bernoulli and Gaussian distributions with DropConnect and DropOut techniques are explored as ensemble methods. All the DropConnect, DropOut and Bayesian layers are applied to the predictor network. Aleatoric uncertainty is modeled by assuming that each data point has an independent uncertainty associated with it. The results of the UQ study are further analyzed to compare the performance of BNN and ensemble methods. Although this study is confined to UQ of fuel property prediction, the methodologies are applicable to other deep learning frameworks that are being widely used in the combustion community.

33 ADVANCED PROPULSION SYSTEMS↗

A Bayesian approach to real-time 3D tumor localization via monoscopic x-ray imaging during treatment delivery

Purpose: Monoscopic x-ray imaging with on-board kV devices is an attractive approach for real-time image guidance in modern radiation therapy such as VMAT or IMRT, but it falls short in providing reliable information along the direction of imaging x-ray. By effectively taking consideration of projection data at prior times and/or angles through a Bayesian formalism, the authors develop an algorithm for real-time and full 3D tumor localization with a single x-ray imager during treatment delivery. Methods: First, a prior probability density function is constructed using the 2D tumor locations on the projection images acquired during patient setup. Whenever an x-ray image is acquired during the treatment delivery, the corresponding 2D tumor location on the imager is used to update the likelihood function. The unresolved third dimension is obtained by maximizing the posterior probability distribution. The algorithm can also be used in a retrospective fashion when all the projection images during the treatment delivery are used for 3D localization purposes. The algorithm does not involve complex optimization of any model parameter and therefore can be used in a ''plug-and-play'' fashion. The authors validated the algorithm using (1) simulated 3D linear and elliptic motion and (2) 3D tumor motion trajectories of a lung and a pancreas patient reproduced by a physical phantom. Continuous kV images were acquired over a full gantry rotation with the Varian TrueBeam on-board imaging system. Three scenarios were considered: fluoroscopic setup, cone beam CT setup, and retrospective analysis. Results: For the simulation study, the RMS 3D localization error is 1.2 and 2.4 mm for the linear and elliptic motions, respectively. For the phantom experiments, the 3D localization error is < 1 mm on average and < 1.5 mm at 95th percentile in the lung and pancreas cases for all three scenarios. The difference in 3D localization error for different scenarios is small and is not statistically significant. Conclusions: The proposed algorithm eliminates the need for any population based model parameters in monoscopic image guided radiotherapy and allows accurate and real-time 3D tumor localization on current standard LINACs with a single x-ray imager.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

BEYONDPLANCK III. Commander3

We describe the computational infrastructure for end-to-end Bayesian cosmic microwave background (CMB) analysis implemented by the BeyondPlanck Collaboration. The code is called Commander3. It provides a statistically consistent framework for global analysis of CMB and microwave observations and may be useful for a wide range of legacy, current, and future experiments. The paper has three main goals. Firstly, we provide a high-level overview of the existing code base, aiming to guide readers who wish to extend and adapt the code according to their own needs or re-implement it from scratch in a different programming language. Secondly, we discuss some critical computational challenges that arise within any global CMB analysis framework, for instance in-memory compression of time-ordered data, fast Fourier transform optimization, and parallelization and load-balancing. Thirdly, we quantify the CPU and RAM requirements for the current BEYONDPLANCK analysis, finding that a total of 1.5 TB of RAM is required for efficient analysis and that the total cost of a full Gibbs sample for LFI is 170 CPU-hrs, including both low-level processing and high-level component separation, which is well within the capabilities of current low-cost computing facilities. The existing code base is made publicly available under a GNU General Public Library (GPL) license.

79 ASTRONOMY AND ASTROPHYSICS↗

The cut-sky cosmic microwave background is not anomalous

The observed angular correlation function of the cosmic microwave background has previously been reported to be anomalous, particularly when measured in regions of the sky uncontaminated by Galactic emission. Recent work by Efstathiou et al. presents a Bayesian comparison of isotropic theories, casting doubt on the significance of the purported anomaly. We extend this analysis to all anisotropic Gaussian theories with vanishing mean (<{delta}T>=0), using the much wider class of models to confirm that the anomaly is not likely to point to new physics. On the other hand if there is any new physics to be gleaned, it results from low-l alignments which will be better quantified by a full-sky statistic. We also consider quadratic maximum likelihood power spectrum estimators that are constructed assuming isotropy. The underlying assumptions are therefore false if the ensemble is anisotropic. Nonetheless we demonstrate that, for theories compatible with the observed sky, these estimators (while no longer optimal) remain statistically superior to pseudo-C{sub l} power spectrum estimators.

79 ASTRONOMY AND ASTROPHYSICS↗

Kernel learning backward SDE filter for data assimilation

In this paper, we develop a kernel learning backward SDE filter method to estimate the state of a stochastic dynamical system based on its partial noisy observations. A system of forward backward stochastic differential equations is used to propagate the state of the target dynamical model, and Bayesian inference is applied to incorporate the observational information. Further, to characterize the dynamical model in the entire state space, we introduce a kernel learning method to learn a continuous global approximation for the conditional probability density function of the target state by using discrete approximated density values as training data. Numerical experiments demonstrate that the kernel learning backward SDE is highly effective.

97 MATHEMATICS AND COMPUTING↗

Uncertainty estimates for the Bayes Inference Engine, (BIE)

In the fall 2007 meeting of the BIB users group, two approaches to making uncertainty estimates were presented. Ken Hanson asserted that if the BFGS optimizer was used, the inverse Hessian matrix was the same as the covariance matrix representing parameter uncertainties. John Pang presented preliminary results of a Monte Carlo method called Randomized Maximum Likelihood (RML). The BFGS/Hessian matrix approach may be applied to the region of the 'ideal model' Approximately 250 parameters describing the object density patches that are varied to match an image of 1,000,000 pixels. I cast this in terms of least squares analysis, as it is much better understood. This not as large a conceptual jump as some suppose because many of the functional blocks in the BIB are taken directly from existing least squares programs. If a Gaussian (normal) probability density function is assumed for both the observation and parameter errors, the Bayesian and least squares result should be identical.

97 MATHEMATICS AND COMPUTING↗

Sensor Selection and Data Validation for Reliable Integrated System Health Management

For new access to space systems with challenging mission requirements, effective implementation of integrated system health management (ISHM) must be available early in the program to support the design of systems that are safe, reliable, highly autonomous. Early ISHM availability is also needed to promote design for affordable operations; increased knowledge of functional health provided by ISHM supports construction of more efficient operations infrastructure. Lack of early ISHM inclusion in the system design process could result in retrofitting health management systems to augment and expand operational and safety requirements; thereby increasing program cost and risk due to increased instrumentation and computational complexity. Having the right sensors generating the required data to perform condition assessment, such as fault detection and isolation, with a high degree of confidence is critical to reliable operation of ISHM. Also, the data being generated by the sensors needs to be qualified to ensure that the assessments made by the ISHM is not based on faulty data. NASA Glenn Research Center has been developing technologies for sensor selection and data validation as part of the FDDR (Fault Detection, Diagnosis, and Response) element of the Upper Stage project of the Ares 1 launch vehicle development. This presentation will provide an overview of the GRC approach to sensor selection and data quality validation and will present recent results from applications that are representative of the complexity of propulsion systems for access to space vehicles. A brief overview of the sensor selection and data quality validation approaches is provided below. The NASA GRC developed Systematic Sensor Selection Strategy (S4) is a model-based procedure for systematically and quantitatively selecting an optimal sensor suite to provide overall health assessment of a host system. S4 can be logically partitioned into three major subdivisions: the knowledge base, the down-select iteration, and the final selection analysis. The knowledge base required for productive use of S4 consists of system design information and heritage experience together with a focus on components with health implications. The sensor suite down-selection is an iterative process for identifying a group of sensors that provide good fault detection and isolation for targeted fault scenarios. In the final selection analysis, a statistical evaluation algorithm provides the final robustness test for each down-selected sensor suite. NASA GRC has developed an approach to sensor data qualification that applies empirical relationships, threshold detection techniques, and Bayesian belief theory to a network of sensors related by physics (i.e., analytical redundancy) in order to identify the failure of a given sensor within the network. This data quality validation approach extends the state-of-the-art, from red-lines and reasonableness checks that flag a sensor after it fails, to include analytical redundancy-based methods that can identify a sensor in the process of failing. The focus of this effort is on understanding the proper application of analytical redundancy-based data qualification methods for onboard use in monitoring Upper Stage sensors.

Garg, Sanjay↗

Capacity Investment under Bayesian Information Updates at Reporting Periods: Model and Application

We consider capacity addition decisions by a new product manufacturer faced with uncertain technology alternatives. The manufacturing capacity addition and technology development occurs in parallel, with preliminary results from a technology project's success providing valuable information to the manufacturer in adding capacity. We solve a stochastic dynamic program with Bayesian updates to obtain the manufacturer's expected profit‐maximizing capacity investment decision. Our model and applications are motivated by the Critical Materials Institute (CMI) (funded by the Department of Energy), which manages research projects focused on mitigating critical material constraints, vital to renewable energy technologies such as direct‐drive wind turbines, electric vehicles, and energy‐efficient lighting. We capture three unique aspects of the problem: first, the learning from project progress depends on task‐based stochastic outcomes and a project's percent‐done. Second, the underlying technology's profitability is based on a model of competition. Third, we evaluate the impact of progress across a portfolio of projects based on a manufacturer's capacity addition. We develop a heuristic that produces results that are close to optimal and can thus be used for large problem sizes. The managerial insights from an application of our model to CMI projects include: (i) technology projects that report the percent‐done of a project earlier increase expected manufacturer profit; (ii) careful choice of “safe bets,” that is, technologies with low profitability but a high probability of success, can increase expected manufacturer profit; (iii) a portfolio of projects can increase profits significantly over separate project evaluation; and (iv) dynamic management of project resources can increase overall profit.

Vedantam, Aditya↗

Utilization of extended bayesian networks in decision making under uncertainty

Bayesian network tool (called IKE for Integrated Knowledge Engine) has been developed to assess the probability of undesirable events. The tool allows indications and observables from sensors and/or intelligence to feed directly into hypotheses of interest, thus allowing one to quantify the probability and uncertainty of these events resulting from very disparate evidence. For example, the probability that a facility is processing nuclear fuel or assembling a weapon can be assessed by examining the processes required, establishing the observables that should be present, then assembling information from intelligence, sensors and other information sources related to the observables. IKE also has the capability to determine tasking plans, that is, prioritize which observable should be collected next to most quickly ascertain the 'true' state and drive the probability toward 'zero' or 'one.' This optimization capability is called 'evidence marshaling.' One example to be discussed is a denied facility monitoring situation; there is concern that certain process(es) are being executed at the site (due to some intelligence or other data). We will show how additional pieces of evidence will then ascertain with some degree of certainty the likelihood of this process(es) as each piece of evidence is obtained. This example shows how both intelligence and sensor data can be incorporated into the analysis. A second example involves real-time perimeter security. For this demonstration we used seismic, acoustic, and optical sensors linked back to IKE. We show how these sensors identified and assessed the likelihood of 'intruder' versus friendly vehicles.

97 MATHEMATICS AND COMPUTING↗

THE LYOT PROJECT DIRECT IMAGING SURVEY OF SUBSTELLAR COMPANIONS: STATISTICAL ANALYSIS AND INFORMATION FROM NONDETECTIONS

The Lyot project used an optimized Lyot coronagraph with extreme adaptive optics at the 3.63 m Advanced Electro-Optical System telescope to observe 86 stars from 2004 to 2007. In this paper, we give an overview of the survey results and a statistical analysis of the observed nondetections around 58 of our targets to place constraints on the population of substellar companions to nearby stars. The observations did not detect any companion in the substellar regime. Since null results can be as important as detections, we analyzed each observation to determine the characteristics of the companions that can be ruled out. For this purpose, we use a Monte Carlo approach to produce artificial companions and determine their detectability by comparison with the sensitivity curve for each star. All the non-detection results are combined using a Bayesian approach and we provide upper limits on the population of giant exoplanets and brown dwarfs for this sample of stars. Our nondetections confirm the rarity of brown dwarfs around solar-like stars and we constrain the frequency of massive substellar companions (M>40 M{sub J}) at orbital separation between and 10 and 50 AU to be {approx_lt}20%.

79 ASTRONOMY AND ASTROPHYSICS↗

Privacy-aware contextual localization using network traffic analysis

The rise of location-based services has enabled many opportunities for content service providers to optimize the content delivery to user's wireless devices based on her location. Since the sharing precise location remains a major privacy concern among the users, certain location-based services rely on contextual location (e.g. residence, work, etc.) as opposed to acquiring user's exact physical location. In this paper, we present PACL (Privacy-Aware Contextual Localizer) model, which can learn user's contextual location just by passively monitoring user's network traffic. PACL can discern a set of vital attributes (statistical and application-based) from user's network traffic, and predict user's contextual location with a very high accuracy. We design and evaluate PACL using real-world network traces of over 1700 users with over 100GB of total data. Our results show that PACL, when built using the Bayesian Network machine learning algorithm, can predict user's contextual location with the accuracy of around 89%.

97 MATHEMATICS AND COMPUTING↗

On the Effects of Cloud Water Content on Passive Microwave Snowfall Retrievals

The Bayesian passive microwave retrievals of snowfall often rely on mathematical matching of the observed vectors of brightness temperature with an a priori database of precipitation profiles and their corresponding brightness temperatures. Mathematical proximity does not necessarily lead to consistent retrievals due to limited information content of passive microwave observations. This paper defines imposter (genuine) vectors of brightness temperature as those that are mathematically close but physically inconsistent (consistent) and characterizes them through the Silhouette Coefficient (SC) analysis. The Neyman–Pearson (NP) hypothesis testing is used to separate the imposter and genuine brightness temperatures based on their associated values of cloud ice (IWP) and liquid water path (LWP), given by coincidences of CloudSat Profiling Radar (CPR) and the Global Precipitation Measurement (GPM) Microwave Imager (GMI). The study determines thresholds for IWP and LWP that allow optimal identification of imposter brightness temperatures of non-snowing and snowing clouds, which can mislead the passive microwave retrieval algorithms to falsely detect or miss the snowfall events. It is demonstrated that emission signal of supercooled liquid water in snowing clouds can lead to improved passive microwave retrieval of snowfall and conditioning the retrievals to the cloud IWP and LWP can result in marginal correction of the snowfall detection probability; however, reduce the probability of false alarm by 6%–8% over sea ice and open oceans.

Snowfall↗

High-throughput spin-bath characterization of spin defects in semiconductors

Detailed knowledge of the local environments of spin defects in semiconductors, such as nitrogenvacancy (NV) centers in diamond or divacancies in silicon carbide, is crucial for optimizing control and entanglement protocols in quantum sensing and information applications. However, at present a direct experimental characterization of individual defect environments is not scalable, as conventional spin-bath measurements are time consuming and difficult to automate. Achieving high-throughput characterization requires short experiments to probe the spin bath. However, with fewer and noisier measurements, the inverse problem of recovering spin-bath properties from measured data becomes ill posed, with multiple spin baths having a high likelihood of yielding the same data. In this work, we present a set of computational tools to resolve the ill-posed inverse problem of recovering the atomic positions and hyperfine couplings of random nuclei surrounding spin defects from sparse, noisy experimental coherence data, which can be obtained in hours. Here, we use a trans-dimensional Bayesian approach that incorporates ab initio data to yield full posterior distributions over nuclear spin environments, enabling robust recovery from limited data. We also provide practical tools and guidelines to determine the limits of detectability for hyperfine couplings under specific dynamical decoupling sequences and sampling conditions. In addition, we demonstrate how the tools developed here, in combination with ab initio simulations of spin baths, can guide the design of efficient experimental protocols for application-specific high-throughput screening. To showcase the utility of our approach, we apply it to design fast dynamical decoupling experiments to characterize the spin baths often individual NV centers in diamond. While the primary focus is on accelerating spin-bath characterization of spin defects, this Bayesian approach also lays the foundation for digital-twin studies of spin defects, where a virtual model of the spin-defect system evolves in real time with ongoing experimental measurements. Together, the set of tools we designed and applied paves the way for scalable deployment of spin defects in semiconductors for quantum sensing and information applications.

Bayesian methods↗