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.
Flapper valve remains open under normal flow conditions but closes upon sudden increases to high rate of flow and remains closed until reset. Valve is fluid/mechanical analog of electrical fuse or circuit breaker. Low-pressure-drop shutoff valve contains flapper machined from cylindrical surface. During normal flow conditions, flapper presents small cross section to flow. (Useful in stopping loss of fluid through leaks in cooling systems.)
Abstract Simulations play a key role for inference in collider physics. We explore various approaches for enhancing the precision of simulations using machine learning, including interventions at the end of the simulation chain (reweighting), at the beginning of the simulation chain (pre-processing), and connections between the end and beginning (latent space refinement). To clearly illustrate our approaches, we use W + jets matrix element surrogate simulations based on normalizing flows as a prototypical example. First, weights in the data space are derived using machine learning classifiers. Then, we pull back the data-space weights to the latent space to produce unweighted examples and employ the Latent Space Refinement ( Laser ) protocol using Hamiltonian Monte Carlo. An alternative approach is an augmented normalizing flow, which allows for different dimensions in the latent and target spaces. These methods are studied for various pre-processing strategies, including a new and general method for massive particles at hadron colliders that is a tweak on the widely-used RamboOnDiet mapping. We find that modified simulations can achieve sub-percent precision across a wide range of phase space.
Here, we present a supervised learning framework of training generative models for density estimation. Generative models, including generative adversarial networks (GANs), normalizing flows, and variational auto-encoders (VAEs), are usually considered as unsupervised learning models, because labeled data are usually unavailable for training. Despite the success of the generative models, there are several issues with the unsupervised training, e.g., requirement of reversible architectures, vanishing gradients, and training instability. To enable supervised learning in generative models, we utilize the score-based diffusion model to generate labeled data. Unlike existing diffusion models that train neural networks to learn the score function, we develop a training-free score estimation method. This approach uses mini-batch-based Monte Carlo estimators to directly approximate the score function at any spatial-temporal location in solving an ordinary differential equation (ODE), corresponding to the reverse-time stochastic differential equation (SDE). This approach can offer both high accuracy and substantial time savings in neural network training. Once the labeled data are generated, we can train a simple, fully connected neural network to learn the generative model in the supervised manner. Compared with existing normalizing flow models, our method does not require the use of reversible neural networks and avoids the computation of the Jacobian matrix. Compared with existing diffusion models, our method does not need to solve the reverse-time SDE to generate new samples. As a result, the sampling efficiency is significantly improved. We demonstrate the performance of our method by applying it to a set of 2D datasets as well as real data from the University of California Irvine (UCI) repository.
We propose a Bayesian physics-informed neural network (B-PINN) to solve both forward and inverse nonlinear problems described by partial differential equations (PDEs) and noisy data. In this Bayesian framework, the Bayesian neural network (BNN) combined with a PINN for PDEs serves as the prior while the Hamiltonian Monte Carlo (HMC) or the variational inference (VI) could serve as an estimator of the posterior. B-PINNs make use of both physical laws and scattered noisy measurements to provide predictions and quantify the aleatoric uncertainty arising from the noisy data in the Bayesian framework. Compared with PINNs, in addition to uncertainty quantification, B-PINNs obtain more accurate predictions in scenarios with large noise due to their capability of avoiding overfitting. We conduct a systematic comparison between the two different approaches for the B-PINNs posterior estimation (i.e., HMC or VI), along with dropout used for quantifying uncertainty in deep neural networks. Our experiments show that HMC is more suitable than VI with mean field Gaussian approximation for the B-PINNs posterior estimation, while dropout employed in PINNs can hardly provide accurate predictions with reasonable uncertainty. Finally, we replace the BNN in the prior with a truncated Karhunen-Loève (KL) expansion combined with HMC or a deep normalizing flow (DNF) model as posterior estimators. The KL is as accurate as BNN and much faster but this framework cannot be easily extended to high-dimensional problems unlike the BNN based framework.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
High-pressure, high-temperature seal flow (leakage) data for nonrotating and rotating Raleigh-step and convergent-tapered-bore seals were characterized in terms of a normalized flow coefficient. The data for normalized Rayleigh-steip and nonrotating tapered-bore seals were in reasonable agreement with theory, but data for the rotating tapered-bore seals were not. The tapered-bore-seal operational clearances estimated from the flow data were significantly larger than calculated. Although clearances are influenced by wear from conical to cylindrical geometry and errors in clearance corrections, the problem was isolated to the shaft temperature - rotational speed clearance correction. The geometric changes support the use of some conical convergence in any seal. Under these conditions rotation reduced the normalized flow coefficiently by nearly 10 percent.
High-pressure, high-temperature seal flow (leakage) data for nonrotating and rotating Raleigh-step and convergent-tapered-bore seals were characterized in terms of a normalized flow coefficient. The data for normalized Rayleigh-step and nonrotating tapered-bore seals were in reasonable agreement with theory, but data for the rotating tapered-bore seals were not. The tapered-bore-seal operational clearances estimated from the flow data were significantly larger than calculated. Although clearances are influenced by wear from conical to cylindrical geometry and errors in clearance corrections, the problem was isolated to the shaft temperature - rotational speed clearance correction. The geometric changes support the use of some conical convergence in any seal. Under these conditions rotation reduced the normalized flow coefficiently by nearly 10 percent.
This report, which is a revision and extension of NACA-TN-1428, presents a compilation of equations, tables, and charts useful in the analysis of high-speed flow of a compressible fluid. The equations provide relations for continuous one-dimensional flow, normal and oblique shock waves, and Prandtl-Meyer expansions for both perfect and imperfect gases. The tables present useful dimensionless ratios for continuous one-dimensional flow and for normal shock waves as functions of Mach number for air considered as a perfect gas. One series of charts presents the characteristics of the flow of air (considered a perfect gas) for oblique shock waves and for cones in a supersonic air stream. A second series shows the effects of caloric imperfections on continuous one-dimensional flow and on the flow through normal and oblique shock waves. (author)
Steady-state pressure drop characteristics of normal hydrogen gas flowing through electrically heated small diameter tube - laminar flow instability
A finite-element model of the Navier-Stokes equations is used for numerical simulation of flow past two normal flat plates arranged side by side at Reynolds number 80 and 160. The results from this simulation indicate that when the gap between the plates is twice the width of a single plate, the individual wakes of the plates behave independently, with the antiphase vortex shedding being dominant. At smaller gap sizes, the in-phase vortex shedding, with strong wake interaction, is favored. The gap flow in those cases becomes biased, with one of the wakes engulfing the other. The direction of the biased flow was found to be switching at irregular intervals, with the time histories of the indicative flow parameters and their power spectra resembling those of a chaotic system.
High-energy physics requires the generation of large numbers of simulated data samples from complex but analytically tractable distributions called matrix elements. Surrogate models, such as normalizing flows, are gaining popularity for this task due to their computational efficiency. We adopt an approach based on flow annealed importance sampling bootstrap (FAB) that evaluates the differentiable target density during training and helps avoid the costly generation of training data in advance. We show that FAB reaches higher sampling efficiency with fewer target evaluations in high dimensions in comparison to other methods.
Not provided.
An arrangement for sensing the fluid separation along a surface which employs a thermally insulating element having a continuous surface blending into and forming a part of the fluid flow surface is described. A sudden decrease in the temperature of the downstream sensor conductor and concomitant increase in the temperature of the upstream sensor conductor is an indication of the separation. When the temperatures are returned to the state achieved during normal flow, the indicator thereby indicates the normal, attached fluid flow. The conductors may be, for example, wires or thin films, and should be within the viscous sub-layer of the expected fluid flow. A single heater and several pairs of sensors and corresponding sensor conductors may be used to detect not only the fluid flow and the separation, but the direction of the fluid flow, over the fluid flow surface.
Nonequilibrium flow field behind normal shock waves in carbon dioxide
Performance and boundary-layer data were taken in a 12 degree 10-inch inlet-diameter conical diffuser of 2:1 exit- to inlet-area ratio. These data were taken for two inlet-boundary-layer conditions. The first condition was that of a thinner inlet boundary later (boundary-layer displacement thickness, delta* approximately equal to 0.034) produced by an inlet section approximately 1 inlet diameter in length between the entrance bell and the diffuser. The second condition was a thicker inlet boundary layer (delta* approximately equal to 0.120) produced by an additional inlet section length of approximately 6 diameters. Longitudinal static-pressure distributions were measured fro wall static orifices. Transverse total- and static-pressure surveys were made at the inlet and exit stations. Boundary-layer velocity distributions were measured at seven stations between the inlet and exit. These data were obtained for a Reynolds number (based on inlet diameter) range of 1 x 10(exp 6) to 3.9 x 10(exp 6). The corresponding Mach number range was from M = 0.2 to choking. At the maximum-power-available condition supersonic flow was obtained as far as 4.5 inches downstream from the diffuser inlet with a maximum Mach number of M approximately equal to 1.5. The total-pressure loss through the diffuser in percentage of inlet dynamic pressure was approximately 2.5 percent for the thinner inlet boundary later and 5.5 percent for the thicker inlet boundary later over the lower subsonic range. These valued increased with increasing flow rate- the values for the thicker inlet boundary later more than those for the thinner inlet boundary layer. The diffuser effectiveness, expressed as the ratio of the actual static-pressure rise to the ideal static-pressure rise, was about 85 percent for the thinner inlet boundary layer and about 67 percent for the thicker inlet boundary later in the lower subsonic range. These values decrease with increasing flow rate. Separated flow was observed for both inlet-boundary-layer conditions in the region of adverse pressure gradient just downstream of the transition curvature from inlet section to diffuser. The flow for the thinner-inlet-boundary-layer condition did not fully re-establish itself along the diffuser walls. The thicker inlet-boundary-layer flow, while not completely re-establishing the normal flow pattern downstream of the separated region, did re-establish more successfully than the thinner inlet boundary layer.