Estimation using sampled-data containing sequentially correlated noise.
Multistage linear dynamic systems with sequentially correlated noise, evaluating filtering, prediction and smoothing procedures
SEARCH · Search NASA
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.
Multistage linear dynamic systems with sequentially correlated noise, evaluating filtering, prediction and smoothing procedures
An experimental investigation of distributed sand-grain surface roughness effects on hypersonic boundary-layer transition and turbulent heating has been performed. Two representative entry vehicle geometries, a sphere-cone aeroshell and a spherical-cap aeroshell, were considered. Cast ceramic models of each geometry were fabricated with distributed sand-grain roughness patterns of different heights that simulated an ablated thermal protection system. Wind tunnel testing was performed at Mach 6 over a range of Reynolds numbers sufficient to produce laminar, transitional, and turbulent flow. Aeroheating and boundary-layer transition onset data were obtained using global phosphor thermography. The experimental heating data are presented herein, as are comparisons to laminar and turbulent smooth-wall heat transfer distributions from computational flow field simulations. The boundary-layer transition data were found to correlate with a functional representation developed in prior roughness studies, although the data scatter was greater owing to the height variability of distributed sand-grain roughness.
A detailed fatigue characterization of Alloy 10, a high strength nickel-based disk alloy, was conducted on test coupons machined from a 'pancake' forging. Smooth bar, strain controlled fatigue testing at various R-ratios was run at representative bore, 750 F, and rim, 1300 F, temperatures. This was followed by notch fatigue testing (Kt=2) run under load control. Analysis of the fatigue data using a Smith-Watson-Topper approach and finite element analysis of the notch root was employed to understand material behavior in these tests. Smooth bar fatigue data showed a significant R-ratio dependence at either test temperature which could be accounted for using a Smith-Watson-Topper parameter (SWT). In general, fatigue life was longer at 750 F than 1300 F for a given SWT. For notch fatigue tests, life was longer at 750 F than 1300 F but only at higher stresses. This was attributed to differences in alloy strength. At lower stresses, finite element analysis suggested that convergence of fatigue life at both temperatures resulted from relaxation of stresses at the notch root in the 1300 F tests.
ABSTRACT There are a number of hypotheses underlying the existence of adversarial examples for classification problems. These include the high‐dimensionality of the data, the high codimension in the ambient space of the data manifolds of interest, and that the structure of machine learning models may encourage classifiers to develop decision boundaries close to data points. This article proposes a new framework for studying adversarial examples that does not depend directly on the distance to the decision boundary. Similarly to the smoothed classifier literature, we define a (natural or adversarial) data point to be ( γ , σ)‐stable if the probability of the same classification is at least for points sampled in a Gaussian neighborhood of the point with a given standard deviation . We focus on studying the differences between persistence metrics along interpolants of natural and adversarial points. We show that adversarial examples have significantly lower persistence than natural examples for large neural networks in the context of the MNIST and ImageNet datasets. We connect this lack of persistence with decision boundary geometry by measuring angles of interpolants with respect to decision boundaries. Finally, we connect this approach with robustness by developing a manifold alignment gradient metric and demonstrating the increase in robustness that can be achieved when training with the addition of this metric.
Understanding the aerodynamic impact of swept-wing ice accretions is a crucial component of the design of modern aircraft. Computer-simulation tools are commonly used to approximate ice shapes, so the necessary level of detail or fidelity of those simulated ice shapes must be understood relative to high-fidelity representations of the ice. Previous tests were performed in the NASA Icing Research Tunnel to acquire high-fidelity ice shapes. Some of those ice shapes are based on aircraft certification requirements. From this database, full-span artificial ice shapes were designed and manufactured for both an 8.9%-scale and 13.3%-scale semispan wing model of the CRM65 which has been established as the full-scale baseline for this swept-wing project. These models were tested in the Walter H. Beech wind tunnel at Wichita State University and at the ONERA (Office national d'etudes et de recherches aerospatiales) F1 facility, respectively. The data collected in the Wichita State University wind tunnel provided a low-Reynolds number baseline study while the pressurized F1 facility produced data over a wide range of Reynolds and Mach numbers with the highest Reynolds number studied being approximately Re = 11.9 by 10 (sup 6). Three different fidelity representations were created based on three different icing conditions. Lower-fidelity ice shapes were created by lofting a smooth ice shape between cross-section cuts of the high-fidelity ice shape. Grit roughness was attached to this smooth ice shape as another fidelity variant. The data indicates that the geometric fidelity of the ice shapes resulted in significant differences in lift and drag. These results were similar at both facilities over the wide range of test conditions utilized.
The acceleration (gravity) anomaly estimates by spacecraft tracking, determined from Doppler residuals, are components of the gravity field directed along the spacecraft Earth line of sight (LOS). These data constitute a set of vector components of a planet's gravity field, the specific component depending upon where the Earth happened to be at the time of each measurement, and they are at varying altitudes above the planet surface. From this data set the gravity field vector components were derived using the method of harmonic splines which imposes a smoothness criterion to select a gravity model compatible with the LOS data. Given the piecewise model it is now possible to upward and downward continue the field quantities desired with a few parameters unlike some other methods which must return to the full dataset for each desired calculation.
Algorithm enables experimental robotic land vehicle to follow automatically route computed on basis of terrain-height-map data. Computer executing algorithm merges coarser global topographical data with finer local topographical data obtained through stereoscopic video system as vehicle moves along, then smoothes merged elevation map, interpolates map to evenly-horizontally-spaced grid points, and differentiates map to produce data on slopes and roughness for use in calculations of traversability and optimum route(s). Distance traveled minimized to extent possible consistent with avoidance of obstacles.
A technique is demonstrated for accelerated stress corrosion testing of high strength aluminum alloys. The method offers better precision and shorter exposure times than traditional pass fail procedures. The approach uses data from tension tests performed on replicate groups of smooth specimens after various lengths of exposure to static stress. The breaking strength measures degradation in the test specimen load carrying ability due to the environmental attack. Analysis of breaking load data by extreme value statistics enables the calculation of survival probabilities and a statistically defined threshold stress applicable to the specific test conditions. A fracture mechanics model is given which quantifies depth of attack in the stress corroded specimen by an effective flaw size calculated from the breaking stress and the material strength and fracture toughness properties. Comparisons are made with experimental results from three tempers of 7075 alloy plate tested by the breaking load method and by traditional tests of statistically loaded smooth tension bars and conventional precracked specimens.
Automatic boundary detection of 4D ultrasound (4DUS) cardiac data is a promising yet challenging application at the intersection of machine learning and medicine. Using recently developed murine 4DUS cardiac imaging data, we demonstrate here a set of three machine learning models that predict left ventricular wall kinematics along both the endo- and epi-cardial boundaries. Each model is fundamentally built on three key features: (1) the projection of raw US data to a lower dimensional subspace, (2) a smoothing spline basis across time, and (3) a strategic parameterization of the left ventricular boundaries. Model 1 is constructed such that boundary predictions are based on individual short-axis images, regardless of their relative position in the ventricle. Model 2 simultaneously incorporates parallel short-axis image data into their predictions. Model 3 builds on the multi-slice approach of model 2, but assists predictions with a single ground-truth position at end-diastole. To assess the performance of each model, Monte Carlo cross validation was used to assess the performance of each model on unseen data. For predicting the radial distance of the endocardium, models 1, 2, and 3 yielded average R2 values of 0.41, 0.49, and 0.71, respectively. Monte Carlo simulations of the endocardial wall showed significantly closer predictions when using model 2 versus model 1 at a rate of 48.67%, and using model 3 versus model 2 at a rate of 83.50%. These finding suggest that a machine learning approach where multi-slice data are simultaneously used as input and predictions are aided by a single user input yields the most robust performance. Subsequently, we explore the how metrics of cardiac kinematics compare between ground-truth contours and predicted boundaries. We observed negligible deviations from ground-truth when using predicted boundaries alone, except in the case of early diastolic strain rate, providing confidence for the use of such machine learning models for rapid and reliable assessments of murine cardiac function. To our knowledge, this is the first application of machine learning to murine left ventricular 4DUS data. Future work will be needed to strengthen both model performance and applicability to different cardiac disease models.
Advances in space robotics technology hinge to a large extent upon the development and deployment of sophisticated new vision-based methods for automated in-space mission operations and scientific survey. To this end, we have developed a new concept for automated terrain analysis that is based upon a generic image enhancement platform|multi-scale retinex (MSR) and visual servo (VS) processing. This pre-conditioning with the MSR and the vs produces a "canonical" visual representation that is largely independent of lighting variations, and exposure errors. Enhanced imagery is then processed with a biologically inspired two-channel edge detection process, followed by a smoothness based criteria for image segmentation. Landing sites can be automatically determined by examining the results of the smoothness-based segmentation which shows those areas in the image that surpass a minimum degree of smoothness. Though the msr has proven to be a very strong enhancement engine, the other elements of the approach|the vs, terrain map generation, and smoothness-based segmentation|are in early stages of development. Experimental results on data from the Mars Global Surveyor show that the imagery can be processed to automatically obtain smooth landing sites. In this paper, we describe the method used to obtain these landing sites, and also examine the smoothness criteria in terms of the imager and scene characteristics. Several examples of applying this method to simulated and real imagery are shown.
Advances in space robotics technology hinge to a large extent upon the development and deployment of sophisticated new vision-based methods for automated in-space mission operations and scientific survey. To this end, we have developed a new concept for automated terrain analysis that is based upon a generic image enhancement platform-multi-scale Retinex (MSR) and visual servo (VS) processing. This pre-conditioning with the MSR and the VS produces a "canonical" visual representation that is largely independent of lighting variations, and exposure errors. Enhanced imagery is then processed with a biologically inspired two-channel edge detection process, followed by a smoothness based criteria for image segmentation. Landing sites can be automatically determined by examining the results of the smoothness-based segmentation which shows those areas in the image that surpass a minimum degree of smoothness. Though the MSR has proven to be a very strong enhancement engine, the other elements of the approach, the VS, terrain map generation, and smoothness-based segmentation, are in early stages of development. Experimental results on data from the Mars Global Surveyor show that the imagery can be processed to automatically obtain smooth landing sites. In this paper, we describe the method used to obtain these landing sites, and also examine the smoothness criteria in terms of the imager and scene characteristics. Several examples of applying this method to simulated and real imagery are shown.
A complete solar resource data set is essential for any stage of a solar energy project - from feasibility studies to daily operations. But measured or modeled solar resource data are prone to data gaps and data quality issues. To mitigate these issues, a data imputation process should be implemented to obtain a complete and reliable temporal and spatial data series. This study focused on imputing temporal scales by applying random and artificial data gaps and then implementing eight imputation methods, including the Kalman filtering and smoothing and stine interpolations. These methods were implemented on 1-minute to half hourly irradiance data for 1 year using a few locations from the National Solar Radiation Database (NSRDB) and ground measurement data set. The results demonstrated that some of the simpler methods, such as the stine and linear interpolation methods, were the relatively best models based on the statistical metrics for imputing NSRDB and ground measurement data, respectively.
Description of experiments undertaken to define the phase-resetting behavior of the circadian rhythm of pupal eclosion in populations of fruitflies. An attempt is made to determine how and why the resetting response depends on the duration of a standard perturbation as well as on the time at which it is given. Plotting a three-dimensional graph of the measured emergence centroids vs the stimulus variables, the data are found to spiral up around a vertical rotation axis. Using a computer, a smooth surface, called the resetting surface, which approximately fits the helicoidal cloud of data points, is obtained and is shown to be best described as a vertical corkscrew linking together tilted planes. This corkscrew feature of the resetting surface is taken to indicate that there is an isolated perturbation following which there is either no circadian rhythm of emergence in the steady state, or one of unpredictable phase. A hypothesis concerning the clock dynamics underlying the eclosion rhythm is briefly sketched which encompasses the main features of known resetting data using single discrete pulses of any perturbing agent.
Empirical models such as the International Reference Ionosphere (IRI) are synthesized from large data bases. They can be viewed as analytical tools to facilitate accessing information stored in the data banks. However, in establishing the models, one has to apply smoothing and averaging procedures that in effect reduce the original information content. This study evaluates the agreement between the data base and the model at two opposite extremes of time resolution. Electron densities and temperatures in the altitude range of 300 to 400 km predicted by the IRI and measured by the AE-C and DE 2 satellites on the level of individual orbits as well as on the level of mission averages are compared. Whereas the averages show excellent agreement, the comparison for individual measurements indicates the limitations of empirical models.
To evaluate the Tropical Rainfall Measuring Mission (TRMM) monthly Ground Validation (GV) rain map, 42 quality controlled tipping bucket rain gauge data (1 minute interpolated rain rates) were utilized. We have compared the gauge data to the surface volumetric rainfall accumulation of NEXRAD reflectivity field, (converting to rain rates using a 0.5 dB resolution smooth Z-R table). The comparison was carried out from data collected at Melbourne, Florida during the month of July 98. GV operational level 3 (L3 monthly) accumulation algorithm was used to obtain surface volumetric accumulations for the radar. The gauge records were accumulated using the 1 minute interpolated rain rates while the radar Volume Scan (VOS) intervals remain less than or equal to 75 minutes. The correlation coefficient for the radar and gauge totals for the monthly time-scale remain at 0.93, however, a large difference was noted between the gauge and radar derived rain accumulation when the radar data interval is either 9 minute, or 10 minute. This difference in radar and gauge accumulation is being explained in terms of the radar scan strategy information. The discrepancy in terms of the Volume Coverage Pattern (VCP) of the NEXRAD is being reported where VCP mode is ascertained using the radar tilt angle information. Hourly radar and gauge accumulations have been computed using the present operational L3 method supplemented with a threshold period of +/- 5 minutes (based on a sensitivity analysis). These radar and gauge accumulations are subsequently improved using a radar hourly scan weighting factor (taking ratio of the radar scan frequency within a time bin to the 7436 total radar scans for the month). This GV procedure is further being improved by introducing a spatial smoothing method to yield reasonable bulk radar to gauge ratio for the hourly and daily scales.
We apply the Trapezoidal/BDF2 (TR/BDF2) temporal discretization scheme to nonlinear grey radiative diffusion. This is a scheme that is not well-known within the radiation transport community, but we show that it offers many desirable characteristics relative to other second-order schemes. Several nonlinear variants of the TR/BDF2 scheme are defined and computationally compared with the Crank-Nicholson scheme. It is found for our test problems that the most accurate TR/BDF2 schemes are those that are fully iterated to nonlinear convergence, but the most efficient TR/BDF2 scheme is one based upon a single Newton iteration. It is also shown that neglecting the contributions to the Jacobian matrix from the cross-sections, which is often done due to a lack of smooth interpolations for tabular cross-section data, has a significant impact upon efficiency.
A lignin-graft-poly(lactic- co -glycolic) acid (PLGA) biopolymer was synthesized with two types of lignin (LGN), alkaline lignin (ALGN) and sodium lignosulfonate (SLGN), at different (A/S)LGN/PLGA ratios (1:2, 1:4, and 1:6 w/w). 1 H NMR and Fourier-transform infrared spectroscopy (FT-IR) confirmed the conjugation of PLGA to LGN. The (A/S)LGN-graft-PLGA biopolymers were used to form nanodelivery systems suitable for entrapment and delivery of drugs for disease treatment. The LGN-graft-PLGA NPs were generally small (100–200 nm), increased in size with the amount of PLGA added, monodisperse, and negatively charged (-48 to -60 mV). Small-angle scattering data showed that particles feature a relatively smooth surface and a compact spherical structure with a distinct core and a shell. The core size and shell thickness varied with the LGN/PLGA ratio, and at a 1:6 ratio, the particles deviated from the core–shell structure to a complex internal structure. The newly developed (A/S)LGN-graft-PLGA NPs are proposed as a potential delivery system for applications in biopharmaceutical, food, and agricultural sectors.
Abstract Molecular dynamics simulation is an important tool in computational materials science and chemistry, and in the past decade it has been revolutionized by machine learning. This rapid progress in machine learning interatomic potentials has produced a number of new architectures in just the past few years. Particularly notable among these are the atomic cluster expansion, which unified many of the earlier ideas around atom-density-based descriptors, and Neural Equivariant Interatomic Potentials (NequIP), a message-passing neural network with equivariant features that exhibited state-of-the-art accuracy at the time. Here we construct a mathematical framework that unifies these models: atomic cluster expansion is extended and recast as one layer of a multi-layer architecture, while the linearized version of NequIP is understood as a particular sparsification of a much larger polynomial model. Our framework also provides a practical tool for systematically probing different choices in this unified design space. An ablation study of NequIP, via a set of experiments looking at in- and out-of-domain accuracy and smooth extrapolation very far from the training data, sheds some light on which design choices are critical to achieving high accuracy. A much-simplified version of NequIP, which we call BOTnet (for body-ordered tensor network), has an interpretable architecture and maintains its accuracy on benchmark datasets.