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.
To evaluate the accuracy and prognostication of the presence of radiologic extranodal extension (rENE) versus pathologic extranodal extension (pENE) in patients with oral cavity squamous cell carcinoma (OSCC).
Thermal fluid processes are inherently multi-physics and multi-scale, involving mass-momentum-energy transport phenomena at multiple scales. Thermal fluid simulation (TFS) is based on solving conservative equations, for which – except for “first-principles” direct numerical simulation – closure relations (CRs) are required to provide microscopic interactions or so-called sub-grid-scale physics. In practice, TFS is realized through reduced-order modeling, and its CRs as low-fidelity models can be informed by observations and data from relevant and adequately evaluated experiments and high-fidelity simulations. This paper is focused on data-driven TFS models, specifically on their development using machine learning (ML). Five ML frameworks are introduced including physics-separated ML (PSML or Type I ML), physics-evaluated ML (PEML or Type II ML), physics-integrated ML (PIML or Type III ML), physics-recovered (PRML or Type IV ML), and physics-discovered ML (PDML or Type V ML). The frameworks vary in their performance for different applications depending on the level of knowledge of governing physics, source, type, amount and quality of available data for training. Notably, outlined for the first time in this paper, Type III models present stringent requirements on modeling, substantial computing resources for training, and high potential in extracting value from “big data” in thermal fluid research. Here the current paper demonstrates and investigates ML frameworks in three examples. First, we utilize the heat diffusion equation with a nonlinear conductivity model formulated by convolutional neural networks (CNNs) and feedforward neural networks (FNNs) to illustrate the applications of Type I, Type II, Type III, and Type V ML. The results indicate a preference for Type II ML under deficient data support. Type III ML can effectively utilize field data, potentially generating more robust predictions than Type I and Type II ML. CNN-based closures exhibit more predictability than FNN-based closures, but CNN-based closures require more training data to obtain accurate predictions. Second, we illustrate how to employ Type I ML and Type II ML frameworks for data-driven turbulence modeling using reference works. Third, we demonstrate Type I ML by building a deep FNN-based slip closure for two-phase flow modeling. The results show that deep FNN-based closures exhibit a bounded error in the prediction domain.
Mojave Desert ecosystem processes are dependent upon the amount and seasonality of precipitation. Multi-decadal periods of drought or above-average rainfall affect landscape vegetation condition, biomass and susceptibility to fire. The seasonality of precipitation events can also affect the likelihood of lightning, a key ignition source for fires. To develop an understanding of precipitation regimes and fire patterns we used monthly average precipitation data and GIS data representing burned areas from 1971-2010. We applied a K-means cluster analysis to the monthly precipitation data identifying three distinct precipitation seasons; winter (October – March), spring (April-June) and summer (July-September) and four discrete precipitation regimes within the Mojave ecoregion.
We present a method for creating spacecraft-like data which can be used to train Machine Learning (ML) models to detect and classify structures in in situ spacecraft data. First, we use the Grad-Shafranov equation to numerically solve for several magnetohydrostatic equilibria which are variations on a known analytic equilibrium. These equilibria are then used as the initial conditions for Particle-In-Cell simulations in which the structures of interest are observed and labeled. We then take one-dimensional slices through the simulations to replicate what a spacecraft collecting data from the simulation would observe. This sliced data then can be used as training data for the initial training of ML models intended for use on spacecraft data. We demonstrate the method applied to the problem of detecting small-scale plasmoids in the magnetotail, which is important for understanding complex magnetotail reconnection dynamics. The simple 1D classifier we train is able to detect more than 70% of the plasmoid points in the data set but also produces a large number of false positives. Our further work on this example problem is detailed, and further potential uses of the method are discussed.
We present a systematic method to classify the four-level system using SU(4) symmetry as the basis group. It is shown that this symmetry allows three dipole transitions which eventually leads to six possible configurations of the four-level system. Using a dressed atom approach, the semi-classical version of each configuration is exactly solved under rotating wave approximation and the symmetry among the Rabi oscillation among various models is studied.
The purpose of this study is to map pyprophyllite distribution at surface of the Nohwa deposit, Korea by using Advanced Spaceborne Thermal Emission and Reflectance Radiometer (ASTER) data. For this, combined Spectral Angle Mapper (SAM), and Matched Filtering (MF) technique based on mathematical algorithm was applied. The regional distribution of high-grade and low-grade pyrophyllite in the Nohwa deposit area could be differentiated by this method. The results of this study show that ASTER data analysis using combination of SAM and MF techniques will assist in exploration of pyrophyllite at the exposed surface.
The unstable periodic orbits (UPOs) embedded in a chaotic attractor after an attractor merging crisis (MC) are classified into three subsets, and employed to reconstruct chaotic saddles in the Kuramoto-Sivashinsky equation. It is shown that in the post-MC regime, the two chaotic saddles evolved from the two coexisting chaotic attractors before crisis can be reconstructed from the UPOs embedded in the pre-MC chaotic attractors. The reconstruction also involves the detection of the mediating UPO responsible for the crisis, and the UPOs created after crisis that fill the gap regions of the chaotic saddles. We show that the gap UPOs originate from saddle-node, period-doubling, and pitchfork bifurcations inside the periodic windows in the post-MC chaotic region of the bifurcation diagram. The chaotic attractor in the post-MC regime is found to be the closure of gap UPOs.
We analyze the theories of gravity modified by a generic nonderivative potential built from the metric, under the minimal requirement of unbroken spatial rotations. Using the canonical analysis, we classify the potentials V according to the number of degrees of freedom (DoF) that propagate at the nonperturbative level. We then compare the nonperturbative results with the perturbative DoF propagating around Minkowski and FRW backgrounds. A generic V implies 6 propagating DoF at the non-perturbative level, with a ghost on Minkowski background. There exist potentials which propagate 5 DoF, as already studied in previous works. Here, no V with unbroken rotational invariance admitting 4 DoF is found. Theories with 3 DoF turn out to be strongly coupled on Minkowski background. Finally, potentials with only the 2 DoF of a massive graviton exist. Their effect on cosmology is simply equivalent to a cosmological constant. Potentials with 2 or 5 DoF and explicit time dependence appear to be a further viable possibility.
One of the goals of the Surveillance, Epidemiology, and End Results (SEER) program is to estimate incidence, prevalence, and mortality of all cancers. To that end, cancer registries across the country maintain a massive database of cancer pathology reports which contain rich information to understand cancer trends. However, these reports are stored in the form of unstructured text, and human annotators are required to read and extract relevant information. In this article, we show that existing deep learning models for automating information extraction from cancer pathology reports can be significantly improved by using ensemble model distillation. We found that by training multiple predictive models and transferring their knowledge to a single, low-resource model, we can reduce the number of highly confident wrong predictions. Our results show that our implemented methods could save 1000s of manual annotation hours.
Explore the source record for details and available documents.
Fault detection and isolation is critical for reliable operation of distribution systems. The ride-through requirements for the distributed energy resources (DER), mandated by the IEEE 1547-2018 standard, makes it challenging to use undervoltage (UV) conditions for fault detection. In addition, with low fault current contribution from these inverter-based DERs, the time-overcurrent relays are also less effective. Thus motivated, this paper presents a learning-based approach for fault detection and localization. A convolutional neural network (CNN)-based model is proposed which uses local voltage and current waveforms from DER locations and feeder substations, for training a zonal classifier. The classifier can be adopted into any relay-like device for discriminating between faults originating from different protection zones. The performance of the proposed approach was tested on publicly available test feeders with distributed photovoltaics (PVs).
A key barrier to applying any smart technology to a building is the requirement of locating and connecting to the necessary resources among the thousands of sensing and control points, i.e., the metadata mapping problem. Existing solutions depend on exhaustive manual annotation of sensor metadata --- a laborious, costly, and hardly scalable process. To reduce the amount of manual effort required, this paper presents a multi-oracle selective sampling framework to leverage noisy labels from information sources with unknown reliability such as existing buildings, which we refer to as weak oracles, for metadata mapping. This framework involves an interactive process, where a small set of sensor instances are progressively selected and labeled for it to learn how to aggregate the noisy labels as well as to predict sensor types. Two key challenges arise in designing the framework, namely, weak oracle reliability estimation and instance selection for querying. To address the first challenge, we develop a clustering-based approach for weak oracle reliability estimation to capitalize on the observation that weak oracles perform differently in different groups of instances. For the second challenge, we propose a disagreement-based query selection strategy to combine the potential effect of a labeled instance on both reducing classifier uncertainty and improving the quality of label aggregation. We evaluate our solution on a large collection of real-world building sensor data from 5 buildings with more than 11,000 sensors of 18 different types. The experiment results validate the effectiveness of our solution, which outperforms a set of state-of-the-art baselines.
Not Available
Here, we compared perennial riparian plant community composition among five a priori hydrogeomorphic channel types at 86 stream reaches in arid ephemeral watersheds. Floristic differences among channel types were assessed using PERMANOVA. GLM were used to identify species and functional groups whose abundance differed among channel types. Abiotic drivers of compositional differences were identified through correlations among biotic and abiotic matrices.
Purpose To determine how consistent the results of different raters are when reviewing the same cases within the Radiation Oncology Incident Learning System (ROILS). Methods Three second-year medical physics graduate students filled out incident reports in spreadsheets set up to mimic ROILS. All students studied the same 33 cases and independently entered their assessments, for a total of 99 reviewed cases. The narratives for these cases were obtained from a published International Commission on Radiological Protection (ICRP) report which included shorter narratives selected from the Radiation Oncology Safety Information System (ROSIS) database. Each category of questions was reviewed to see how consistent the results were by utilizing free-marginal multirater kappa analysis. The percentage of cases where all raters shared full agreement or full disagreement was recorded to show which questions were answered consistently by multiple raters for a given case. The consistency among the raters was analyzed between ICRP and ROSIS cases to see if either group led to more reliable results. Results The categories where all raters agreed 100 percent in their choices were the event type (93.94 percent of cases 0.946 kappa) and the likelihood of the event being harmful to the patient (42.42 percent of cases 0.409 kappa). The categories where all raters disagreed 100 percent in their choices were the dosimetric severity scale (39.39 percent of cases 0.139 kappa) and the potential future toxicity (48.48 percent of cases 0.205 kappa). ROSIS had more cases where all raters disagreed than ICRP (23.06 percent of cases compared to 15.58 percent, respectively). Conclusion Despite reviewing the same cases, the results among the three raters was widespread. ROSIS narratives were shorter than ICRP, which suggests that longer narratives lead to more consistent results. This study shows that the incident reporting system can be optimized to yield more consistent results.
Purpose: To quantify the impact of 4D PET/CT on PERCIST metrics in lung and liver tumors in NSCLC and colorectal cancer patients. Methods: 32 patients presenting lung or liver tumors of 1–3 cm size affected by respiratory motion were scanned on a GE Discovery 690 PET/CT. The bed position with lesion(s) affected by motion was acquired in a 12 minute PET LIST mode and unlisted into 8 bins with respiratory gating. Three different CT maps were used for attenuation correction: a clinical helical CT (CT-clin), an average CT (CT-ave), and an 8-phase 4D CINE CT (CT-cine). All reconstructions were 3D OSEM, 2 iterations, 24 subsets, 6.4 Gaussian filtration, 192×192 matrix, non-TOF, and non-PSF. Reconstructions using CT-clin and CT-ave used only 3 out of the 12 minutes of the data (clinical protocol); all 12 minutes were used for the CT-cine reconstruction. The percent change of SUVbw-peak and SUVbw-max was calculated between PET-CTclin and PET-CTave. The same percent change was also calculated between PET-CTclin and PET-CTcine in each of the 8 bins and in the average of all bins. A 30% difference from PET-CTclin classified lesions as progressive metabolic disease (PMD) using maximum bin value and the average of eight bin values. Results: 30 lesions in 25 patients were evaluated. Using the bin with maximum SUVbw-peak and SUVbw-max difference, 4 and 13 lesions were classified as PMD, respectively. Using the average bin values for SUVbw-peak and SUVbw-max, 3 and 6 lesions were classified as PMD, respectively. Using PET-CTave values for SUVbw-peak and SUVbw-max, 4 and 3 lesions were classified as PMD, respectively. Conclusion: These results suggest that response evaluation in 4D PET/CT is dependent on SUV measurement (SUVpeak vs. SUVmax), number of bins (single or average), and the CT map used for attenuation correction.
The energy states of the exciton in a single square GaAs-based quantum well are obtained from a numerical solution of the three-dimensional Schrödinger equation for the envelope of the exciton wave function. This equation is based on the exciton effective energy operator with a spherical approximation of the Luttinger Hamiltonian. The calculated states are classified based on the types of one-dimensional functions for the factorized form of the wave function. The upper limit for the energies of the exciton states in a quantum well is confirmed by the complex scaling method.