Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Parameter Estimation”

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 1,405 records · Page 78

Cosmological parameters from large scale structure - geometric versus shape information

The matter power spectrum as derived from large scale structure (LSS) surveys contains two important and distinct pieces of information: an overall smooth shape and the imprint of baryon acoustic oscillations (BAO). We investigate the separate impact of these two types of information on cosmological parameter estimation for current data, and show that for the simplest cosmological models, the broad-band shape information currently contained in the SDSS DR7 halo power spectrum (HPS) is by far superseded by geometric information derived from the baryonic features. An immediate corollary is that contrary to popular beliefs, the upper limit on the neutrino mass m{sub ν} presently derived from LSS combined with cosmic microwave background (CMB) data does not in fact arise from the possible small-scale power suppression due to neutrino free-streaming, if we limit the model framework to minimal ΛCDM+m{sub ν}. However, in more complicated models, such as those extended with extra light degrees of freedom and a dark energy equation of state parameter w differing from -1, shape information becomes crucial for the resolution of parameter degeneracies. This conclusion will remain true even when data from the Planck spacecraft are combined with SDSS DR7 data. In the course of our analysis, we update both the BAO likelihood function by including an exact numerical calculation of the time of decoupling, as well as the HPS likelihood, by introducing a new dewiggling procedure that generalises the previous approach to models with an arbitrary sound horizon at decoupling. These changes allow a consistent application of the BAO and HPS data sets to a much wider class of models, including the ones considered in this work. All the cases considered here are compatible with the conservative 95%-bounds Σm{sub ν} < 1.16eV, N{sub eff} = 4.8±2.0.

79 ASTRONOMY AND ASTROPHYSICS↗

GVIPS Models and Software

Two reports discuss, respectively, (1) the generalized viscoplasticity with potential structure (GVIPS) class of mathematical models and (2) the Constitutive Material Parameter Estimator (COMPARE) computer program. GVIPS models are constructed within a thermodynamics- and potential-based theoretical framework, wherein one uses internal state variables and derives constitutive equations for both the reversible (elastic) and the irreversible (viscoplastic) behaviors of materials. Because of the underlying potential structure, GVIPS models not only capture a variety of material behaviors but also are very computationally efficient. COMPARE comprises (1) an analysis core and (2) a C++-language subprogram that implements a Windows-based graphical user interface (GUI) for controlling the core. The GUI relieves the user of the sometimes tedious task of preparing data for the analysis core, freeing the user to concentrate on the task of fitting experimental data and ultimately obtaining a set of material parameters. The analysis core consists of three modules: one for GVIPS material models, an analysis module containing a specialized finite-element solution algorithm, and an optimization module. COMPARE solves the problem of finding GVIPS material parameters in the manner of a design-optimization problem in which the parameters are the design variables.

Arnold, Steven M.↗

Controls for space structures

Assembly and operation of large space structures (LSS) in orbit will require robot-assisted docking and berthing of partially-assembled structures. These operations require new solutions to the problems of controls. This is true because of large transient and persistent disturbances, controller-structure interaction with unmodeled modes, poorly known structure parameters, slow actuator/sensor dynamical behavior, and excitation of nonlinear structure vibrations during control and assembly. For on-orbit assembly, controllers must start with finite element models of LSS and adapt on line to the best operating points, without compromising stability. This is not easy to do, since there are often unmodeled dynamic interactions between the controller and the structure. The indirect adaptive controllers are based on parameter estimation. Due to the large number of modes in LSS, this approach leads to very high-order control schemes with consequent poor stability and performance. In contrast, direct model reference adaptive controllers operate to force the LSS to track the desirable behavior of a chosen model. These schemes produce simple control algorithms which are easy to implement on line. One problem with their use for LSS has been that the model must be the same dimension as the LSS - i.e., quite large. A control theory based on the command generator tracker (CGT) ideas of Sobel, Mabins, Kaufman and Wen, Balas to obtain very low-order models based on adaptive algorithms was developed. Closed-loop stability for both finite element models and distributed parameter models of LSS was proved. In addition, successful numerical simulations on several LSS databases were obtained. An adaptive controller based on our theory was also implemented on a flexible robotic manipulator at Martin Marietta Astronautics. Computation schemes for controller-structure interaction with unmodeled modes, the residual mode filters or RMF, were developed. The RMF theory was modified to compensate slow actuator/sensor dynamics. These new ideas are being applied to LSS simulations to demonstrate the ease with which one can incorporate slow actuator/sensor effects into our design. It was also shown that residual mode filter compensation can be modified for small nonlinearities to produce exponentially stable closed-loop control. A theory for disturbance accommodating controllers based on reduced order models of structures was developed, and stability results for these controllers in closed-loop with large-scale finite element models of structures were obtained.

Balas, Mark↗

Estimating short-period dynamics using an extended Kalman filter

An extended Kalman filter (EKF) is used to estimate the parameters of a low-order model from aircraft transient response data. The low-order model is a state space model derived from the short-period approximation of the longitudinal aircraft dynamics. The model corresponds to the pitch rate to stick force transfer function currently used in flying qualities analysis. Because of the model chosen, handling qualities information is also obtained. The parameters are estimated from flight data as well as from a six-degree-of-freedom, nonlinear simulation of the aircraft. These two estimates are then compared and the discrepancies noted. The low-order model is able to satisfactorily match both flight data and simulation data from a high-order computer simulation. The parameters obtained from the EKF analysis of flight data are compared to those obtained using frequency response analysis of the flight data. Time delays and damping ratios are compared and are in agreement. This technique demonstrates the potential to determine, in near real time, the extent of differences between computer models and the actual aircraft. Precise knowledge of these differences can help to determine the flying qualities of a test aircraft and lead to more efficient envelope expansion.

Bauer, Jeffrey E.↗

ACTION-SPACE CLUSTERING OF TIDAL STREAMS TO INFER THE GALACTIC POTENTIAL

We present a new method for constraining the Milky Way halo gravitational potential by simultaneously fitting multiple tidal streams. This method requires three-dimensional positions and velocities for all stars to be fit, but does not require identification of any specific stream or determination of stream membership for any star. We exploit the principle that the action distribution of stream stars is most clustered when the potential used to calculate the actions is closest to the true potential. Clustering is quantified with the Kullback-Leibler Divergence (KLD), which also provides conditional uncertainties for our parameter estimates. We show, for toy Gaia-like data in a spherical isochrone potential, that maximizing the KLD of the action distribution relative to a smoother distribution recovers the input potential. The precision depends on the observational errors and number of streams; using K III giants as tracers, we measure the enclosed mass at the average radius of the sample stars accurate to 3% and precise to 20%-40%. Recovery of the scale radius is precise to 25%, biased 50% high by the small galactocentric distance range of stars in our mock sample (1-25 kpc, or about three scale radii, with mean 6.5 kpc). 20-25 streams with at least 100 stars each are required for a stable confidence interval. With radial velocities (RVs) to 100 kpc, all parameters are determined with ∼10% accuracy and 20% precision (1.3% accuracy for the enclosed mass), underlining the need to complete the RV catalog for faint halo stars observed by Gaia.

79 ASTRONOMY AND ASTROPHYSICS↗

The GMRT archive atomic gas survey – II. Mass modelling and dark matter halo properties across late-type spirals

Studying the kinematics and mass modelling of galaxies from H i 21 cm data provides valuable insights into the properties of both the baryonic components and the dark matter halo in nearby galaxies. Despite many observational studies, mass modelling of galaxies remains challenging due to different limitations. For example, most of the previous studies involving mass modelling are based on rotation curves derived from 2D velocity fields from H i or H α spectroscopic observation which are often affected by beam smearing and projection effect. However, kinematic modelling done by fitting the ‘Tilted ring model’ to 3D data cube is not affected by these issues. In this study, we present and compare 3D kinematic modelling of a pilot sample of 11 galaxies from the GMRT archive atomic gas survey (GARCIA) using two different publicly available pipelines. We model the observed H i rotation curve using 3.6-μm infrared data and SDSS r-band data for stellar contribution, H i surface density profile for gas, and Navarro–Frenk–White profile for dark matter halo; and employ the Markov chain Monte Carlo optimization method for parameter estimation. Further, to validate our analysis, we revisit important scaling relations, e.g. the M gas –M star relation, M star –M halo relation, M gas –M halo relation and Baryonic Tully–Fisher relation. Here, the scaling relations from our analysis are broadly consistent with that reported in the literature. A larger sample of galaxies from GARCIA in the near future will allow studying these scaling relations in greater details.

79 ASTRONOMY AND ASTROPHYSICS↗

RICO: A NEW APPROACH FOR FAST AND ACCURATE REPRESENTATION OF THE COSMOLOGICAL RECOMBINATION HISTORY

We present RICO, a code designed to compute the ionization fraction of the universe during the epoch of hydrogen and helium recombination with an unprecedented combination of speed and accuracy. This is accomplished by training the machine learning code PICO on the calculations of a multilevel cosmological recombination code which self-consistently includes several physical processes that were neglected previously. After training, RICO is used to fit the free electron fraction as a function of the cosmological parameters. While, for example, at low redshifts (z {approx}< 900), much of the net change in the ionization fraction can be captured by lowering the hydrogen fudge factor in RECFAST by about 3%, RICO provides a means of effectively using the accurate ionization history of the full recombination code in the standard cosmological parameter estimation framework without the need to add new or refined fudge factors or functions to a simple recombination model. Within the new approach presented here, it is easy to update RICO whenever a more accurate full recombination code becomes available. Once trained, RICO computes the cosmological ionization history with negligible fitting error in {approx}10 ms, a speedup of at least 10{sup 6} over the full recombination code that was used here. Also RICO is able to reproduce the ionization history of the full code to a level well below 0.1%, thereby ensuring that the theoretical power spectra of cosmic microwave background (CMB) fluctuations can be computed to sufficient accuracy and speed for analysis from upcoming CMB experiments like Planck. Furthermore, it will enable cross-checking different recombination codes across cosmological parameter space, a comparison that will be very important in order to assure the accurate interpretation of future CMB data.

79 ASTRONOMY AND ASTROPHYSICS↗

A parallel workflow implementation for PEST version 13.6 in high-performance computing for WRF-Hydro version 5.0: a case study over the midwestern United States

The Weather Research and Forecasting Hydrological (WRF-Hydro) system is a state-of-the-art numerical model that models the entire hydrological cycle based on physical principles. As with other hydrological models, WRF-Hydro parameterizes many physical processes. Hence, WRF-Hydro needs to be calibrated to optimize its output with respect to observations for the application region. When applied to a relatively large domain, both WRF-Hydro simulations and calibrations require intensive computing resources and are best performed on multimode, multicore high-performance computing (HPC) systems. Typically, each physics-based model requires a calibration process that works specifically with that model and is not transferrable to a different process or model. The parameter estimation tool (PEST) is a flexible and generic calibration tool that can be used in principle to calibrate any of these models. In its existing configuration, however, PEST is not designed to work on the current generation of massively parallel HPC clusters. To address this issue, we ported the parallel PEST to HPCs and adapted it to work with WRF-Hydro. The porting involved writing scripts to modify the workflow for different workload managers and job schedulers, as well as to connect the parallel PEST to WRF-Hydro. To test the operational feasibility and the computational benefits of this first-of-its-kind HPC-enabled parallel PEST, we developed a case study using a flood in the midwestern United States in 2013. Results on a problem involving the calibration of 22 parameters show that on the same computing resources used for parallel WRF-Hydro, the HPC-enabled parallel PEST can speed up the calibration process by a factor of up to 15 compared with commonly used PEST in sequential mode. The speedup factor is expected to be greater with a larger calibration problem (e.g., more parameters to be calibrated or a larger size of study area).

54 ENVIRONMENTAL SCIENCES↗

Image phase shift invariance based cloud motion displacement vector calculation method for ultra-short-term solar PV power forecasting

Irradiance received on the earth's surface is the main factor that affects the output power of solar PV plants, and is chiefly determined by the cloud distribution seen in a ground-based sky image at the corresponding moment in time. It is the foundation for those linear extrapolation-based ultra-short-term solar PV power forecasting approaches to obtain the cloud distribution in future sky images from the accurate calculation of cloud motion displacement vectors (CMDVs) by using historical sky images. Theoretically, the CMDV can be obtained from the coordinate of the peak pulse calculated from a Fourier phase correlation theory (FPCT) method through the frequency domain information of sky images. The peak pulse is significant and unique only when the cloud deformation between two consecutive sky images is slight enough, which is likely possible for a very short time interval (such as 1?min or shorter) with common changes in the speed of cloud. Sometimes, there will be more than one pulse with similar values when the deformation of the clouds between two consecutive sky images is comparatively obvious under fast changing cloud speeds. This would probably lead to significant errors if the CMDVs were still only obtained from the single coordinate of the peak value pulse. However, the deformation estimation of clouds between two images and its influence on FPCT-based CMDV calculations are terrifically complex and difficult because the motion of clouds is complicated to describe and model. Therefore, to improve the accuracy and reliability under these circumstances in a simple manner, an image-phase-shift-invariance (IPSI) based CMDV calculation method using FPCT is proposed for minute time scale solar power forecasting. First, multiple different CMDVs are calculated from the corresponding consecutive images pairs obtained through different synchronous rotation angles compared to the original images by using the FPCT method. Second, the final CMDV is generated from all of the calculated CMDVs through a centroid iteration strategy based on its density and distance distribution. Third, the influence of different rotation angle resolution on the final CMDV is analyzed as a means of parameter estimation. Simulations under various scenarios including both thick and thin clouds conditions indicated that the proposed IPSI-based CMDV calculation method using FPCT is more accurate and reliable than the original FPCT method, optimal flow (OF) method, and particle image velocimetry (PIV) method.

14 SOLAR ENERGY↗

Field-Scale Lysimeter Studies of Low-Activity Waste Form Degradation

Before vitrified and cementitious nuclear waste can be disposed of at the U.S. Department of Energy’s Hanford Site Integrated Disposal Facility (IDF), a performance assessment (PA) must be conducted. A field lysimeter is one method capable of confirming that laboratory experiments provide reasonable model parameter estimates for PA release models. Field lysimeters are also useful in demonstrating that release models can accurately predict material degradation in a field environment. This report describes the implementation plan for a study of LAW waste form degradation using the Field Lysimeter Test Facility (FLTF) on the Hanford Site. This facility provides an opportunity to study glass and cementitious waste forms that will be produced during Hanford cleanup activities in field-relevant conditions. Results from this study will: 1) provide field experimental data on degradation of various waste forms; 2) be used to refine process model descriptions of contaminant (source term) release from the waste forms; 3) reduce uncertainties about the representativeness of laboratory testing results for determining long-term waste form performance under field conditions; and 4) improve confidence in the IDF PA by providing data that verify parameters and assumptions used in the PA modeling. It is expected that this study will lead to improved stakeholder confidence in the safe disposal of LAW at the IDF and aid in the adoption of informed facility designs with the potential to reduce operational costs.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Computational control of flexible aerospace systems

The main objective of this project is to establish a distributed parameter modeling technique for structural analysis, parameter estimation, vibration suppression and control synthesis of large flexible aerospace structures. This report concentrates on the research outputs produced in the last two years. The main accomplishments can be summarized as follows. A new version of the PDEMOD Code had been completed based on several incomplete versions. The verification of the code had been conducted by comparing the results with those examples for which the exact theoretical solutions can be obtained. The theoretical background of the package and the verification examples has been reported in a technical paper submitted to the Joint Applied Mechanics & Material Conference, ASME. A brief USER'S MANUAL had been compiled, which includes three parts: (1) Input data preparation; (2) Explanation of the Subroutines; and (3) Specification of control variables. Meanwhile, a theoretical investigation of the NASA MSFC two-dimensional ground-based manipulator facility by using distributed parameter modeling technique has been conducted. A new mathematical treatment for dynamic analysis and control of large flexible manipulator systems has been conceived, which may provide an embryonic form of a more sophisticated mathematical model for future modified versions of the PDEMOD Codes.

Sharpe, Lonnie, Jr.↗

Residence time distributions in surface transient storage zones in streams: Estimation via signal deconvolution

Little is known about the impact of surface transient storage (STS) zones on reach-scale transport and the fate of dissolved nutrients in streams. Exchange with these locations may influence the rates of nutrient cycling often observed in whole-stream tracer experiments, particularly because they are sites of organic matter collection and lower flow velocities than those observed in the thalweg. We performed a conservative stream tracer experiment (slug of dissolved NaCl) in the Ipswich River in northeastern Massachusetts and collected solute tracer data both in the thalweg and adjacent STS zones at three locations in a fifth-order reach. Tracer time series observed in STS zones are an aggregate of residence time distributions (RTDs) of the upstream transport to that point (RTDTHAL) and that of the temporary storage within these zones (RTDSTS). Here we demonstrate the separation of these two RTDs to determine the RTDSTS specifically. Total residence times for these individual STS zones range from 4.5 to 7.5 h, suggesting that these zones have the potential to host important biogeochemical transformations in stream systems. All of the RTDSTS show substantial deviations from the ideal prescribed by the two-state (mobile/immobile) mass transfer equations. The deviations indicate a model mismatch and that parameter estimation based on the mass transfer equations will yield misleading values.

58 GEOSCIENCES↗

Feasibility Study of SDAS Instrumentation's Ability to Identify Mobile Launcher (ML)/Crawler-Transporter (CT) Modes During Rollout Operations

The Space Launch System (SLS) and its Mobile Launcher (ML) will be transported to the launch pad via the Crawler-Transporter (CT) system. Rollout (i.e., transportation) loads produce structural loads on the integrated SLS/Orion Multi-Purpose Crew Vehicle (MPCV) launch vehicle which are of a concern with respect to fatigue. As part of the risk reduction process and in addition to the modal building block test approach that has been adopted by the SLS Program, acceleration data will be obtained during rollout for use in modal parameter estimation. There are several occurrences where the ML/CT will be transported either into the Vertical Assembly Building (VAB) or to the launch pad and back without the SLS stack as part of the Kennedy Space Center (KSC) Exploration Ground Systems (EGS) Integrated Test and Checkout (ITCO). NASA KSC EGS has instrumentation installed on both the ML and CT to record data during rollout, at the launch pad, and during liftoff. The EGS instrumentation on the ML, which includes accelerometers, is referred to as the Sensor Data Acquisition System (SDAS). The EGS instrumentation on the CT, which also includes accelerometers, is referred to as the CT Data Acquisition System (CTDAS). The forces and accelerations applied to the ML and CT during a rollout event will be higher than any of the planned building block modal tests. This can be very beneficial in helping identify nonlinear behavior in the structure. Developing modal parameters from the same test hardware in multiple boundary conditions and under multiple levels of excitation is a key step in developing a well correlated FEM. The purpose of this study was three fold. First, determine the target modes of the ML/CT in its rollout configuration. Second, determine if the test degrees of freedom (DOF) corresponding to the layout of the SDAS/CTDAS accelerometers (i.e. position and orientation) is sufficient to identify the target modes. Third, determine if the Generic Rollout Forcing Functions (GRFF's) is sufficient for identifying the ML/CT target modes accounting for variations in CT speed, modal damping, and sensor/ambient background noise levels. The finding from the first part of this study identified 28 target modes of the ML/CT rollout configuration based upon Modal Effective Mass Fractions (MEFF) and engineering judgement. The finding from the second part of this study showed that the SDAS/CTDAS accelerometers (i.e. position and orientation) are able to identify a sufficient number of the target modes to support model correlation of the ML/CT FEM. The finding from the third part of this study confirms the GRFFs sufficiently excite the ML/CT such that varying quantities of the defined target modes should be able to be extracted when utilizing an Experimental Modal Analysis (EMA) Multi-Input Multi-Output (MIMO) analysis approach. An EMA analysis approach was used because Operational Modal Analysis (OMA) tools were not available and the GRFFs were sufficiently uncorrelated. Two key findings from this third part of the study are that the CT speed does not show a significant impact on the ability to extract the modal parameters and that keeping the ambient background noise observed at each accelerometer location at or below 30 µgrms is essential to the success of this approach.

Winkel, James P.↗

Breaking the degeneracy between polarization efficiency and cosmological parameters in CMB experiments

Accurate cosmological parameter estimates using polarization data of the cosmic microwave background (CMB) put stringent requirements on map calibration, as highlighted in the recent results from the Planck satellite. In this paper, we point out that a model-dependent determination of polarization calibration can be achieved by the joint fit of the temperature-E-mode cross-power spectrum (TE) and E-mode auto-power spectrum (EE). This provides a valuable cross-check to band-averaged polarization efficiency measurements determined using other approaches. We demonstrate that, in ΛCDM, the combination of the TE and EE constrain polarization calibration with sub-percent uncertainty with Planck data and 2% uncertainty with sptpol data. We arrive at similar conclusions when extending ΛCDM to include the amplitude of lensing A L , the number of relativistic species N eff , or the sum of the neutrino masses Σm ν . Furthermore, the uncertainties on cosmological parameters are minimally impacted when marginalizing over polarization calibration, except, as can be expected, for the uncertainty on the amplitude of the primordial scalar power spectrum ln(10 10 A s ), which increases by 20–50%. However, this information can be fully recovered by adding temperature auto-power spectrum (TT) information. For current and future ground-based experiments, SPT-3G and CMB-S4, we forecast the cosmological parameter uncertainties to be minimally degraded when marginalizing over polarization calibration parameters. In addition, CMB-S4 could constrain its polarization calibration at the level of ~0.2% by combining TE and EE, and reach ~0.06% by also including TT. We therefore conclude that relying on calibrating against Planck polarization maps, whose statistical uncertainty is limited to ~0.5%, would be insufficient for upcoming experiments.

79 ASTRONOMY AND ASTROPHYSICS↗

FY2022 Progress on Imbibition Testing in Containment Science

Estimation of two-phase fluid flow properties is important to understand and predict water and gas movement through the vadose zone for agricultural, hydrogeological, and engineering applications, such as for vapor-phase contaminant transport and/or containment of noble gases in the subsurface. In this second progress report of FY22, we present two ongoing activities related to imbibition testing on volcanic rock samples. We present the development of a new analytical solution predicting the temperature response observed during imbibition into dry samples, as discussed in our previous first progress report for FY22. We also illustrate the use of a multi-modal capillary pressure distribution to simulate both early- and late-time imbibition data collected on tuff core that can exhibit multiple pore types. These FY22 imbibition tests were conducted for an extended period (i.e., far beyond the time required for the wetting front to reach the top of the sample), which is necessary for parameter estimation and characterization of two different pore types within the samples.

42 ENGINEERING↗

Catalog-level blinding on the bispectrum for DESI-like galaxy surveys

We evaluate the performance of the catalog-level blind analysis technique (blinding) presented in Brieden et al. (2020) in the context of a fixed template power spectrum and bispectrum analysis. This blinding scheme, which is tailored for galaxy redshift surveys similar to the Dark Energy Spectroscopic Instrument (DESI), has two components: the so-called “AP blinding” (concerning the dilation parameters α$_{∥}$, α$_{⊥}$) and “RSD blinding” (redshift space distortions, affecting the growth rate parameter f). Through extensive testing, including checks for the RSD part in cubic boxes, the impact of AP blinding on mocks with realistic survey sky coverage, and the implementation of a full AP+RSD blinding pipeline, our analysis demonstrates the effectiveness of the technique in preserving the integrity of cosmological parameter estimation when the analysis includes the bispectrum statistic. We emphasize the critical role of sophisticated — and difficult to accidentally unblind — blinding methods in precision cosmology.

79 ASTRONOMY AND ASTROPHYSICS↗

ECLEIRS: Exact conservation law embedded identification of reduced states for parameterized nonlinear conservation laws from sparse and noisy data

Multi-query applications such as parameter estimation, uncertainty quantification and design optimization for parameterized partial differential equation (PDE) systems are expensive. While reduced/latent state dynamics approaches for parameterized PDEs offer a viable alternative, these approaches rely on high-quality data and struggle with highly sparse spatiotemporal noisy measurements typically obtained from experiments. Furthermore, there is no guarantee that these models satisfy governing physical conservation laws. In this article, we propose a reduced state dynamics approach, referred to as ECLEIRS, that embeds exact conservation in the solution and flux representation by utilizing a space-time divergence-free neural network formulation. We compare ECLEIRS with other reduced state dynamics approaches, those that do not enforce any physical constraints and those with physics-informed loss functions, for three shock-propagation problems: 1-D advection, 1-D Burgers and 2-D Euler equations. In conclusion, the numerical experiments conducted in this study demonstrate that ECLEIRS provides the most accurate prediction of dynamics for unseen parameters even in the presence of highly sparse and noisy data.

97 MATHEMATICS AND COMPUTING↗