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.
Presented is a solution for recovering the wavefront and an extended object. It builds upon the VSM architecture and deconvolution algorithms. Simulations are shown for recovering the wavefront and extended object from noisy data.
Abstract Quantum kernel methods show promise for accelerating data analysis by efficiently learning relationships between input data points that have been encoded into an exponentially large Hilbert space. While this technique has been used successfully in small-scale experiments on synthetic datasets, the practical challenges of scaling to large circuits on noisy hardware have not been thoroughly addressed. Here, we present our findings from experimentally implementing a quantum kernel classifier on real high-dimensional data taken from the domain of cosmology using Google’s universal quantum processor, Sycamore. We construct a circuit ansatz that preserves kernel magnitudes that typically otherwise vanish due to an exponentially growing Hilbert space, and implement error mitigation specific to the task of computing quantum kernels on near-term hardware. Our experiment utilizes 17 qubits to classify uncompressed 67 dimensional data resulting in classification accuracy on a test set that is comparable to noiseless simulation.
This data package includes data and scripts from the manuscript “Denoising autoencoder for reconstructing sensor observation data and predicting evapotranspiration: noisy and missing values repair and uncertainty quantification”.The study addressed common challenges faced in environmental sensing and modeling, including uncertain input data, missing sensor observations, and high-dimensional datasets with interrelated but redundant variables. Point-scaled meteorological and soil sensor observations were perturbed with noises and missing values, and denoising autoencoder (DAE) neural networks were developed to reconstruct the perturbed data and further predict evapotranspiration. This study concluded that (1) the reconstruction quality of each variable depends on its cross-correlation and alignment to the underlying data structure, (2) uncertainties from the models were overall stronger than those from the data corruption, and (3) there was a tradeoff between reducing bias and reducing variance when evaluating the uncertainty of the machine learning models.This package includes:(1) Four ipython scripts (.ipynb): “DAE_train.ipynb” trains and evaluates DAE neural networks, “DAE_predict.ipynb” makes predictions from the trained DAE models, “ET_train.ipynb” trains and evaluates ET prediction neural networks, and “ET_predict.ipynb” makes predictions from trained ET models.(2) One python file (.py): “methods.py” includes all user-defined functions and python codes used in the ipython scripts.(3) A “sub_models” folder that includes five trained DAE neural networks (in pytorch format, .pt), which could be used to ingest input data before being fed to the downstream ET models in ‘ET_train.ipynb” or ‘ET_predict.ipynb’.(4) Two data files (.csv). Daily meteorological, vegetation, and soil data is in “df_data.csv”, where “df_meta.csv” contains the location and time information of “df_data.csv”. Each row (index) in “df_meta.csv” corresponds to each row in “df_data.csv”. These data files are formatted to follow the data structure requirements and be directly used in the ipython scripts, and they have been shuffled chronologically to train machine learning models. The meteorological and soil data was collected using point sensors between 2019-2023 at(4.a) Three shrub-dominated field sites in East River, Colorado (named “ph1”, “ph2” and “sg5” in “df_meta.csv”, where “ph1” and “ph2” were located at PumpHouse Hillslopes, and “sg5” was at Snodgrass Mountain meadow) and(4.b) One outdoor, mesoscale, and herbaceous-dominated experiment in Berkeley, California (named “tb” in “df_meta.csv”, short for Smartsoils Testbed at Lawrence Berkeley National Lab).- See "df_data_dd.csv" and "df_meta_dd.csv" for variable descriptions and the Methods section for additional data processing steps. See "flmd.csv" and "README.txt" for brief file descriptions.- All ipython scripts and python files are written in and require PYTHON language software.
Here, we present a likelihood-free probabilistic inversion method based on normalizing flows for high-dimensional inverse problems. The proposed method is composed of two complementary networks: a summary network for data compression and an inference network for parameter estimation. The summary network encodes raw observations into a fixed-size vector of summary features, while the inference network generates samples of the approximate posterior distribution of the model parameters based on these summary features. The posterior samples are produced in a deep generative fashion by sampling from a latent Gaussian distribution and passing these samples through an invertible transformation. We construct this invertible transformation by sequentially alternating conditional invertible neural network and conditional neural spline flow layers. The summary and inference networks are trained simultaneously. We apply the proposed method to an inversion problem in groundwater hydrology to estimate the posterior distribution of the log-conductivity field conditioned on spatially sparse time-series observations of the system’s hydraulic head responses. The conductivity field is represented with 706 degrees of freedom in the considered problem. Comparison with the likelihood-based iterative ensemble smoother PEST-IES method demonstrates that the proposed method accurately estimates the parameter posterior distribution and the observations’ predictive posterior distribution at a fraction of the inference time of PEST-IES.
Conventional diesel combustion, also known as Mixing-Controlled Compression Ignition (MCCI), is expected to be the primary power source for medium- and heavy-duty vehicles for decades to come. Displacing petroleum-based ultra-low-sulfur diesel (ULSD) as much as possible with low-net-carbon biofuels will become necessary to help mitigate effects on climate change. Neat biofuels may have difficulty meeting current diesel fuel standards but blends of 30% biofuel in ULSD show potential as ‘drop-in’ fuels. These blends must not make significant changes to the combustion phasing of the MCCI process if they are to be used interchangeably with neat ULSD. An important aspect of MCCI phasing is the ignition delay (ID), i.e. the time between the start of fuel injection and the initial premixed autoignition that initiates the MCCI process. Bench experiments can evaluate the expected ignition delay of a fuel via cetane number (CN) or alternative methods such as the indicated cetane number (ICN); however, neither CN nor ICN correlate perfectly with the ignition delay measured in actual engine experiments. Furthermore, there is no standardized methodology on how to quantify MCCI ignition delay from engine cylinder pressure measurements, creating difficulties in cross-study comparison. In this study, several engine ignition delay calculation methods are evaluated for robustness in deriving ignition delay on both a cycle-averaged and cycle-to-cycle basis. Eight biofuel blends with varying ICN, oxygen concentration and other fuel properties were used to study the different methods. This yields a thorough analysis of how certain biofuel blends affect ignition delay and the entire MCCI process, as well as a thorough evaluation of the differences between the ID calculation methods. Many of these methods are equally valid, but the choice of method has a significant impact on the resulting ID, which must be carefully considered when evaluating results across multiple studies.
Conventional diesel combustion, also known as Mixing-Controlled Compression Ignition (MCCI), is expected to be the primary power source for medium- and heavy-duty vehicles for decades to come. Displacing petroleum-based ultra-low-sulfur diesel (ULSD) as much as possible with low-net-carbon biofuels will become necessary to help mitigate effects on climate change. Neat biofuels may have difficulty meeting current diesel fuel standards but blends of 30% biofuel in ULSD show potential as ‘drop-in’ fuels. These blends must not make significant changes to the combustion phasing of the MCCI process if they are to be used interchangeably with neat ULSD. An important aspect of MCCI phasing is the ignition delay (ID), i.e. the time between the start of fuel injection and the initial premixed autoignition that initiates the MCCI process. Bench experiments can evaluate the expected ignition delay of a fuel via cetane number (CN) or alternative methods such as the indicated cetane number (ICN); however, neither CN nor ICN correlate perfectly with the ignition delay measured in actual engine experiments. Furthermore, there is no standardized methodology on how to quantify MCCI ignition delay from engine cylinder pressure measurements, creating difficulties in cross-study comparison. In this study, several engine ignition delay calculation methods are evaluated for robustness in deriving ignition delay on both a cycle-averaged and cycle-to-cycle basis. Eight biofuel blends with varying ICN, oxygen concentration and other fuel properties were used to study the different methods. This yields a thorough analysis of how certain biofuel blends affect ignition delay and the entire MCCI process, as well as a thorough evaluation of the differences between the ID calculation methods. Many of these methods are equally valid, but the choice of method has a significant impact on the resulting ID, which must be carefully considered when evaluating results across multiple studies.
School presentation for work done related to ADAPD during 2021 MSTS summer student program
Modeling the environment is one of the most crucial issues for the development and research of autonomous robot and tele-perception. Though the physical robot operates (navigates and performs various tasks) in the real world, any type of reasoning, such as situation assessment, planning or reasoning about action, is performed based on information in its internal world. Hence, the robot's intentional actions are inherently constrained by the models it has. These models may serve as interfaces between sensing modules and reasoning modules, or in the case of telerobots serve as interface between the human operator and the distant robot. A robot operating in a known restricted environment may have a priori knowledge of its whole possible work domain, which will be assimilated in its World Model. As the information in the World Model is relatively fixed, an Environment Model must be introduced to cope with the changes in the environment and to allow exploring entirely new domains. Introduced here is an algorithm that uses dense range data collected at various positions in the environment to refine and update or generate a 3-D volumetric model of an environment. The model, which is intended for autonomous robot navigation and tele-perception, consists of cubic voxels with the possible attributes: Void, Full, and Unknown. Experimental results from simulations of range data in synthetic environments are given. The quality of the results show great promise for dealing with noisy input data. The performance measures for the algorithm are defined, and quantitative results for noisy data and positional uncertainty are presented.
Self-organizing maps are neural-network techniques for representing noisy, multidimensional data aligned along a lower-dimensional and nonlinear manifold. For a large set of noisy data, each element of a finite set of codebook vectors is iteratively moved in the direction of the data closest to the winner codebook vector. Through successive iterations, the codebook vectors begin to align with the trends of the higher-dimensional data. Prior investigations of ice shapes have focused on using self-organizing maps to characterize mean ice forms. The Icing Research Branch has recently acquired a high resolution three dimensional scanner system capable of resolving ice shape surface roughness. A method is presented for the evaluation of surface roughness variations using high-resolution surface scans based on a self-organizing map representation of the mean ice shape. The new method is demonstrated for 1) an 18-in. NACA 23012 airfoil 2 AOA just after the initial ice coverage of the leading 5 of the suction surface of the airfoil, 2) a 21-in. NACA 0012 at 0AOA following coverage of the leading 10 of the airfoil surface, and 3) a cold-soaked 21-in.NACA 0012 airfoil without ice. The SOM method resulted in descriptions of the statistical coverage limits and a quantitative representation of early stages of ice roughness formation on the airfoils. Limitations of the SOM method are explored, and the uncertainty limits of the method are investigated using the non-iced NACA 0012 airfoil measurements.
Selecting suitable landing sites is fundamental to achieving many mission objectives in planetary robotic lander missions. However, due to sensing limitations, landing sites which are both safe and scientifically valuable often cannot be determined reliably from orbit, particularly, in icy moon missions where orbital sensing data is noisy and incomplete. This paper presents an active perception approach to Entry Descent and Landing (EDL) which enables the lander to autonomously plan informative descent trajectories, acquire high quality sensing data during descent and exploit this additional information to select higher utility landing sites. Our approach consists of two components: probabilistic modeling of landing site features and approximate trajectory planning using a sampling based planner. The proposed framework allows the lander to plan long horizons paths and remain robust to noisy data. Results in simulated environments show large performance improvements over alternative approaches and show promise that our approach has strong potential to improve science return of not only icy moon missions but EDL systems in general.
Predicting vehicle mass is critical to accurately estimate energy use and emissions of commercial trucks. However, data from vehicle telematics is often not at sufficient temporal resolution or accuracy for use in model-based detection methods. In this work, a new statistical mass prediction technique is described for heavy-duty vehicles that incorporates the use Gaussian Belief Propagation (GBP) for probabilistic inference. Similar to Bayesian inference models, the GBP model typically requires less labeled training data than other contemporary machine learning techniques. First, a factor graph is constructed, and a set of Gaussian belief nodes with associated means and variances are fitted to the training data. To better handle noisy input data, the GBP mass prediction model utilizes a k-nearest factors (kNF) algorithm for probabilistic inference on unseen testing data. The proposed method is compared with a classical weighted k-nearest neighbors (kNN) regressor. This statistical kNF-GBP model works even with low-quantity, low-quality initial training data, while being capable of realtime mass estimation. Unlike the kNN regressor, the GBP model produces a measure of uncertainty with its predictions. The proposed method is validated using curve-sampled driving data collected from multiple cloud-connected Class 8 regional haul diesel trucks. Both the kNN regressor and the kNF-GBP mass prediction model were able to predict payload mass with coefficients of determination above 0.97 with minimal data preprocessing.
Machine learning techniques are applied to the NASA Langley Research Center's expansive database of helicopter noise measurements containing over 1500 steady flight conditions for ten different helicopters. These techniques are then used to develop models capable of predicting the operating conditions under which significant Blade-Vortex Interaction noise will be generated for any conventional helicopter. A measure for quantifying the overall ground noise exposure of a particular helicopter operating condition is developed. This measure is then used to classify the measured flight conditions as noisy or not-noisy. These data are then parameterized on a nondimensional basis that defines the main rotor operating condition and are then scaled to remove bias. Several machine learning methods are then applied to these data. The developed models show good accuracy in identifying the noisy operating region for helicopters not included in the training data set. Noisy regions are accurately identified for a variety of different helicopters. One of these models is applied to estimate changes in the noisy operating region as vehicle drag and ambient atmospheric conditions are varied.
In this work, we develop a new Bayesian framework based on deep neural networks to be able to extrapolate in space-time using historical data and to quantify uncertainties arising from both noisy and gappy data in physical problems. Specifically, the proposed approach has two stages: (1) prior learning and (2) posterior estimation. At the first stage, we employ the physics-informed Generative Adversarial Networks (PI-GAN) to learn a functional prior either from a prescribed function distribution, e.g., Gaussian process, or from historical data and physics. At the second stage, we employ the Hamiltonian Monte Carlo (HMC) method to estimate the posterior in the latent space of PI-GANs. In addition, we use two different approaches to encode the physics: (1) automatic differentiation, used in the physicsinformed neural networks (PINNs) for scenarios with explicitly known partial differential equations (PDEs), and (2) operator regression using the deep operator network (DeepONet) for PDE-agnostic scenarios. We then test the proposed method for (1) meta-learning for one-dimensional regression, and forward/inverse PDE problems (combined with PINNs); (2) PDE-agnostic physical problems (combined with DeepONet), e.g., fractional diffusion as well as saturated stochastic (100-dimensional) flows in heterogeneous porous media; and (3) spatial-temporal regression problems, i.e., inference of a marine riser displacement field using experimental data from the Norwegian Deepwater Programme (NDP). The results demonstrate that the proposed approach can provide accurate predictions as well as uncertainty quantification given very limited scattered and noisy data, since historical data could be available to provide informative priors. In summary, the proposed method is capable of learning flexible functional priors, e.g., both Gaussian and non-Gaussian process, and can be readily extended to big data problems by enabling mini-batch training using stochastic HMC or normalizing flows since the latent space is generally characterized as low dimensional.
Ultrasonic testing (UT) for nondestructive evaluation (NDE) is a critical entity necessary to resolve both the quality and precision questions of complex parts evolving from the classical approaches to the innovative additive manufacturing (AM) process. This modality provides the essential quantitative information for acceptance and potential flaw detectionof apart under investigation. A primary ingredient in UT besides the required precision robotic hardware for theacquisition of high-qualitymeasurement data is the underlying signal processing. It is here that much of the system performance capability resides. In this report,we discuss the basic steps in UT signal processing along with current capabilities that must be achieved in order to satisfy the criticaldemands of the various LLNL programs.In itsbasic foundations, signal processing is essentially the “extraction of critical information (signals) from noisy, uncertain data.” Clearly, in a perfect world the best signal processing is none---just make a reliable, noise-free, uncertainty-free, measurement. Unfortunately, even the best of systems is still straddled within the confines and limitations of physical instrumentation. With this in mind, we discuss the development of signal processing techniques toextract the desired UT information from noisy measurement data. We start with a discussion of a simple homogeneous representation of a “part” under investigation and its insonification by an ultrasonicexcitation from a physics-based perspective. Once developed, we briefly discussthe various choices of excitation signals and their tradeoffs.Next,we discuss a simulation approach employing simple models from the signal processing perspective and then move on to the development of the basic signal processing approach to ultrasonic signal processing all based on estimating the pulse arrival estimation. Starting with simple peak detection techniques, progressing to the processing workhorse so-called “matched-filter” and finally to the more sophisticated “model-based matched filter.”The underlying pre-and post-processing of noisy measurement data along with the application of these techniques are subsequently demonstrated on experimental ultrasonic data.Finally, we present some results for processing “weld” data when the ultrasonic excitation signal is not directly available.
A method for characterizing ice shapes using a self-organizing map (SOM) technique is presented. Self-organizing maps are neural-network techniques for representing noisy, multi-dimensional data aligned along a lower-dimensional and possibly nonlinear manifold. For a large set of noisy data, each element of a finite set of codebook vectors is iteratively moved in the direction of the data closest to the winner codebook vector. Through successive iterations, the codebook vectors begin to align with the trends of the higher-dimensional data. In information processing, the intent of SOM methods is to transmit the codebook vectors, which contains far fewer elements and requires much less memory or bandwidth, than the original noisy data set. When applied to airfoil ice accretion shapes, the properties of the codebook vectors and the statistical nature of the SOM methods allows for a quantitative comparison of experimentally measured mean or average ice shapes to ice shapes predicted using computer codes such as LEWICE. The nature of the codebook vectors also enables grid generation and surface roughness descriptions for use with the discrete-element roughness approach. In the present study, SOM characterizations are applied to a rime ice shape, a glaze ice shape at an angle of attack, a bi-modal glaze ice shape, and a multi-horn glaze ice shape. Improvements and future explorations will be discussed.
Abstract Motivation Circadian rhythms are approximately 24-h endogenous cycles that control many biological functions. To identify these rhythms, biological samples are taken over circadian time and analyzed using a single omics type, such as transcriptomics or proteomics. By comparing data from these single omics approaches, it has been shown that transcriptional rhythms are not necessarily conserved at the protein level, implying extensive circadian post-transcriptional regulation. However, as proteomics methods are known to be noisier than transcriptomic methods, this suggests that previously identified arrhythmic proteins with rhythmic transcripts could have been missed due to noise and may not be due to post-transcriptional regulation. Results To determine if one can use information from less-noisy transcriptomic data to inform rhythms in more-noisy proteomic data, and thus more accurately identify rhythms in the proteome, we have created the Multi-Omics Selection with Amplitude Independent Criteria (MOSAIC) application. MOSAIC combines model selection and joint modeling of multiple omics types to recover significant circadian and non-circadian trends. Using both synthetic data and proteomic data from Neurospora crassa, we showed that MOSAIC accurately recovers circadian rhythms at higher rates in not only the proteome but the transcriptome as well, outperforming existing methods for rhythm identification. In addition, by quantifying non-circadian trends in addition to circadian trends in data, our methodology allowed for the recognition of the diversity of circadian regulation as compared to non-circadian regulation. Availability and implementation MOSAIC’s full interface is available at https://github.com/delosh653/MOSAIC. An R package for this functionality, mosaic.find, can be downloaded at https://CRAN.R-project.org/package=mosaic.find. Supplementary information Supplementary data are available at Bioinformatics online.
Distribution-based analyses, such as the distribution of relaxation times (DRT) and the distribution of diffusion times (DDT), present model-free alternatives to equivalent circuit modeling for analysis of electrochemical impedance spectroscopy (EIS) data. However, reconstructing such distributions from noisy impedance data is an ill-posed problem that must be solved with specialized inversion algorithms, requiring careful control and tuning. Furthermore, most inversion algorithms developed to date can only solve problems of limited complexity. Herein, we present a new hierarchical Bayesian method for EIS inversion, leveraging efficient algorithms for optimization and Hamiltonian Monte Carlo (HMC) sampling to solve models of arbitrary complexity. We overcome the challenge of ad-hoc parameter tuning by encoding intrinsic characteristics of the DRT and DDT into flexible prior distributions and “pre-calibrating” the model to simulated data. This approach is versatile, highly robust to noise, and provides quantitative estimates of both the error structure of the data and the uncertainty in the recovered distributions. The model is validated with simulated data to demonstrate accurate recovery of the DRT and the DDT. The method also shows promise for simultaneous recovery of multiple distributions, raising the intriguing possibility of semi-autonomous EIS analysis and ad-hoc model construction. Finally, the practical utility of the method is illustrated with experimental data. Throughout, we draw comparisons to several recently published EIS inversion methodologies.
We report simulating and predicting multiscale problems that couple multiple physics and dynamics across many orders of spatiotemporal scales is a great challenge that has not been investigated systematically by deep neural networks (DNNs). Herein, we develop a framework based on operator regression, the so-called deep operator network (DeepONet), with the long-term objective to simplify multiscale modeling by avoiding the fragile and time-consuming “hand-shaking” interface algorithms for stitching together heterogeneous descriptions of multiscale phenomena. To this end, as a first step, we investigate if a DeepONet can learn the dynamics of different scale regimes, one at the deterministic macroscale and the other at the stochastic microscale regime with inherent thermal fluctuations. Specifically, we test the effectiveness and accuracy of the DeepONet in predicting multirate bubble growth dynamics, which is described by a Rayleigh–Plesset (R–P) equation at the macroscale and modeled as a stochastic nucleation and cavitation process at the microscale by dissipative particle dynamics (DPD). First, we generate data using the R–P equation for multirate bubble growth dynamics caused by randomly time-varying liquid pressures drawn from Gaussian random fields (GRFs). Our results show that properly trained DeepONets can accurately predict the macroscale bubble growth dynamics and can outperform long short-term memory networks. We also demonstrate that the DeepONet can extrapolate accurately outside the input distribution using only very few new measurements. Subsequently, we train the DeepONet with DPD data corresponding to stochastic bubble growth dynamics. Although the DPD data are noisy and we only collect sparse data points on the trajectories, the trained DeepONet model is able to predict accurately the mean bubble dynamics for time-varying GRF pressures. Taken together, our findings demonstrate that DeepONets can be employed to unify the macroscale and microscale models of the multirate bubble growth problem, hence providing new insight into the role of operator regression via DNNs in tackling realistic multiscale problems and in simplifying modeling with heterogeneous descriptions.