Search NASA⌕ Search

SEARCH · Search NASA

Results for “timing error evaluation”

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 649 records · Page 36

A laboratory system for the investigation of rain fade compensation techniques for Ka-band satellites

The design and performance of a rain fade simulation/counteraction system on a laboratory simulated 30/20 GHz, time division multiple access (TDMA) satellite communications testbed is evaluated. Severe rain attenuation of electromagnetic radiation at 30/20 GHz occurs due to the carrier wavelength approaching the water droplet size. Rain in the downlink path lowers the signal power present at the receiver, resulting in a higher number of bit errors induced in the digital ground terminal. The laboratory simulation performed at NASA Lewis Research Center uses a programmable PIN diode attenuator to simulate 20 GHz satellite downlink geographic rain fade profiles. A computer based network control system monitors the downlink power and informs the network of any power threshold violations, which then prompts the network to issue commands that temporarily increase the gain of the satellite based traveling wave tube (TWT) amplifier. After the rain subsides, the network returns the TWT to the normal energy conserving power mode. Bit error rate (BER) data taken at the receiving ground terminal serves as a measure of the severity of rain degradation, and also evaluates the extent to which the network can improve the faded channel.

Svoboda, James S.↗

Planar Doppler Velocimetry for Large-Scale Wind Tunnel Testing

Recently, Planar Doppler Velocimetry (PDV) has been shown by several laboratories to offer an attractive means for measuring three-dimensional velocity vectors everywhere in a light sheet placed in a flow. Unlike other optical means of measuring flow velocities, PDV is particularly attractive for use in large wind tunnels where distances to the sample region may be several meters, because it does not require the spatial resolution and tracking of individual scattering particles or the alignment of crossed beams at large distances. To date, demonstrations of PDV have been made either in low speed flows without quantitative comparison to other measurements, or in supersonic flows where the Doppler shift is large and its measurement is relatively insensitive to instrumental errors. Moreover, most reported applications have relied on the use of continuous-wave lasers, which limit the measurement to time-averaged velocity fields. This work summarizes the results of two previous studies of PDV in which the use of pulsed lasers to obtain instantaneous velocity vector fields is evaluated. The objective has been to quantitatively define and demonstrate PDV capabilities for applications in large-scale wind tunnels that are intended primarily for the production testing of subsonic aircraft. For such applications, the adequate resolution of low-speed flow fields requires accurate measurements of small Doppler shifts that are obtained at distances of several meters from the sample region. The use of pulsed lasers provides the unique capability to obtain not only time-averaged fields, but also their statistical fluctuation amplitudes and the spatial excursions of unsteady flow regions such as wakes and separations. To accomplish the objectives indicated, the PDV measurement process is first modeled and its performance evaluated computationally. The noise sources considered include those related to the optical and electronic properties of Charge-Coupled Device (CCD) arrays and to speckle effects associated with coherent illumination from pulsed lasers. The signal noise estimates are incorporated into the PDV signal analysis process and combined with computed scattering signals using a Mie scattering theory for polydisperse smoke particles. The relevant parameters incorporate a range of practical aerodynamic test conditions and facility sizes. The results define the optimum instrument configurations, show that the expected signal levels from a practical PDV system are sufficiently large to allow its useful application in large facilities, and show that the expected velocity measurement uncertainties are small compared to the mean velocities of interest for most subsonic, large-scale wind tunnel testing. Experimental studies using several experimental bench-top setups are then described that validate the physics of the PDV model and to calibrate its computed results. The validated model allows estimates of the uncertainties of PDV measurements and a complete definition of the PDV capabilities to be made with sufficient confidence to decide the viability of PDV for large-scale wind tunnel applications.

McKenzie, Robert L.↗

Measurement Capabilities of Planar Doppler Velocimetry in Large-Scale Wind Tunnels

Over the past few years, Planar Doppler Velocimetry (PDT) has been shown by several laboratories to offer an attractive means for measuring three-dimensional velocity vectors everywhere in a light sheet placed in a flow. Unlike some other optical means of measuring flow velocities, PDT is particularly attractive for use in large wind tunnels where distances to the sample region may be several meters, because it does not require the spatial resolution and tracking of individual scattering particles or the alignment of crossed beams at large distances. To date, demonstrations of PDT (also called Doppler Global Velocimetry by some authors) have been made either in low speed flows without quantitative comparison to other measurements, or in supersonic flows where the Doppler shift is large and its measurement is relatively insensitive to instrumental errors. Moreover, most reported applications have relied on the use of continuous-wave lasers, which limit the measurement to time-averaged velocity fields. This work summarizes the results of two previous studies of PDT in which the use of pulsed lasers to obtain instantaneous velocity vector fields is evaluated. The objective has been to quantitatively define and demonstrate PDT capabilities for applications in large-scale wind tunnels that are intended primarily for the testing of rotorcraft and subsonic aircraft at speeds typically less than 100 m/s. For such applications, the adequate resolution of low-speed flow fields requires accurate measurements of small Doppler shifts that are obtained at distances of several meters from the sample region and with a field of view that is sufficient to encompass the entire region of interest. The use of pulsed lasers provides the unique capability to obtain not only time-averaged fields, but also their statistical fluctuation amplitudes and the spatial excursions of unsteady flow regions such as wakes, separations, and rotor-tip vortices. To accomplish the objectives of these studies, the PDT measurement process was first modeled and its performance evaluated computationally. The noise sources considered included those related to the optical and electronic properties of Charge-Coupled Device (CCD) arrays and to speckle effects associated with the coherent illumination of aerosols from pulsed lasers. The signal noise estimates were incorporated into the PDT signal analysis process and combined with computed scattering signals using a Mae scattering theory for a distributed range of aerosol particle sizes. The results are used to define the necessary instrument configuration, to show that the expected signal levels from a practical PDT system are sufficiently large to allow its useful application in large facilities, and to show that the expected velocity measurement uncertainties are small compared to the mean velocities of interest for most subsonic, large-scale wind tunnel testing, Experimental studies using several bench-top setups are then described that validate the physics of the PDT model and demonstrate the ability to obtain accurate PDT measurements using procedures that are compatible with large-scale wind tunnel operations. The PDT measurement capabilities found in this study lead to the conclusion that PDT offers significant advantages compared to other means of measuring velocity fields in large-scale wind tunnels, including those operating at low speeds.

McKenzie, Robert L.↗

A Comparison of Two Balance Calibration Model Building Methods

Simulated strain-gage balance calibration data is used to compare the accuracy of two balance calibration model building methods for different noise environments and calibration experiment designs. The first building method obtains a math model for the analysis of balance calibration data after applying a candidate math model search algorithm to the calibration data set. The second building method uses stepwise regression analysis in order to construct a model for the analysis. Four balance calibration data sets were simulated in order to compare the accuracy of the two math model building methods. The simulated data sets were prepared using the traditional One Factor At a Time (OFAT) technique and the Modern Design of Experiments (MDOE) approach. Random and systematic errors were introduced in the simulated calibration data sets in order to study their influence on the math model building methods. Residuals of the fitted calibration responses and other statistical metrics were compared in order to evaluate the calibration models developed with different combinations of noise environment, experiment design, and model building method. Overall, predicted math models and residuals of both math model building methods show very good agreement. Significant differences in model quality were attributable to noise environment, experiment design, and their interaction. Generally, the addition of systematic error significantly degraded the quality of calibration models developed from OFAT data by either method, but MDOE experiment designs were more robust with respect to the introduction of a systematic component of the unexplained variance.

DeLoach, Richard↗

Normalized sensitivities and parameter identifiability of in situ diffusion experiments on Callovo-Oxfordian clay at Bure site

DIR (Diffusion of Inert and Reactive tracers) is an experimental program performed by ANDRA at Bure underground research laboratory in Meuse/Haute Marne (France) to characterize diffusion and retention of radionuclides in Callovo-Oxfordian (C-Ox) argillite. In situ diffusion experiments were performed in vertical boreholes to determine diffusion and retention parameters of selected radionuclides. C-Ox clay exhibits a mild diffusion anisotropy due to stratification. Interpretation of in situ diffusion experiments is complicated by several non-ideal effects caused by the presence of a sintered filter, a gap between the filter and borehole wall and an excavation disturbed zone (EdZ). The relevance of such non-ideal effects and their impact on estimated clay parameters have been evaluated with numerical sensitivity analyses and synthetic experiments having similar parameters and geometric characteristics as real DIR experiments. Normalized dimensionless sensitivities of tracer concentrations at the test interval have been computed numerically. Tracer concentrations are found to be sensitive to all key parameters. Sensitivities are tracer dependent and vary with time. These sensitivities are useful to identify which are the parameters that can be estimated with less uncertainty and find the times at which tracer concentrations begin to be sensitive to each parameter. Synthetic experiments generated with prescribed known parameters have been interpreted automatically with INVERSE-CORE{sup 2D} and used to evaluate the relevance of non-ideal effects and ascertain parameter identifiability in the presence of random measurement errors. Identifiability analysis of synthetic experiments reveals that data noise makes difficult the estimation of clay parameters. Parameters of clay and EdZ cannot be estimated simultaneously from noisy data. Models without an EdZ fail to reproduce synthetic data. Proper interpretation of in situ diffusion experiments requires accounting for filter, gap and EdZ. Estimates of the effective diffusion coefficient and the porosity of clay are highly correlated, indicating that these parameters cannot be estimated simultaneously. Accurate estimation of D{sub e} and porosities of clay and EdZ is only possible when the standard deviation of random noise is less than 0.01. Small errors in the volume of the circulation system do not affect clay parameter estimates. Normalized sensitivities as well as the identifiability analysis of synthetic experiments have provided additional insight on inverse estimation of in situ diffusion experiments and will be of great benefit for the interpretation of real DIR in situ diffusion experiments.

54 ENVIRONMENTAL SCIENCES↗

Regulatory Decision-Making Process Using 'Weight of Evidence: An Evidence Integration Approach' by the State of Washington at the Hanford Site - 20163

Performance Assessment and Environmental Impact Statement Modeling, Human and Ecological Risk Assessments are the tools used by the Washington State Department of Ecology (Ecology) to make regulatory decisions for permitting and cleanup. These efforts all inherently have large amounts of uncertainty no matter how many alternative approaches are attempted. Professional feedback from tribal governments, stakeholders, federal and state regulators, the public and independent peer review groups are taken into account in this process of informed decision making. The process also includes full evaluation of the comments and the associated reviews made by the various peer review groups such as the Low -Level Waste Disposal Facility Federal Review Group (DOE order 435.1) and the Nuclear Regulatory Commission (NRC). The evolution of technology has produced quicker and more efficient fate and transport modeling tools with related quantification of uncertainties and their significance. Groundwater and vadose zone models have been expanded across the entire Hanford site. Many of the technical approaches identified in the Washington State Regulations are updated using current knowledge to better understand the complexities of the Hanford Site using the latest state of the art tools for risk assessment and modeling. More emphasis and time are directed on defining uncertainty and then managing the uncertainty. One added tool is to run sensitivity cases based on informed decision making. The weight of evidence approach considers all relevant information in an integrative assessment that takes into account the kinds of evidence available, the quality and quantity, the uncertainty and error associated with each type, and how they fit together. Evidence of integration involves looking at systematic reviews and evaluating their strengths and weaknesses. In conclusion, Ecology, Washington State's lead regulatory agency at the Hanford site, has had to modify its site closure approach compared to the rest of the state. A more holistic approach dealing with the Hanford site as an integrated system has evolved leading to a defensible informed approach for regulatory closure decisions. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Feasibility of Correlation-Aware Inference and Universal Precision Scaling in Bonse–Hart Ultra-Small-Angle Neutron Scattering

Bonse–Hart ultra-small-angle neutron scattering (USANS) provides access to micrometre-scale structure, but useful measurements often require long counting times. In this work, we test whether the expected smoothness of the scattering profile can be exploited to improve data quality at lower counting statistics. We apply a Gaussian-process-based method to Bonse–Hart USANS data and evaluate its performance on pseudo-measurements generated from high-statistics experiments under Poisson statistics. This provides a stringent test of how well the underlying I(Q) profile can be reconstructed when the available counts are substantially reduced. We further show that, in the counting-limited regime, the reconstruction error follows a universal scaling behaviour that differs from the usual independent-counting expectation. At higher counts, the improvement crosses over to a resolution-limited regime set by analyser-angle discretization and rocking-curve width. These results clarify when correlation-aware inference is useful in USANS and provide a practical basis for improving measurement efficiency and beam-time usage.

Tung, Chi-Huan [ORNL] (ORCID:0000000221972074)↗

Near Real-Time Assessment of Anatomic and Dosimetric Variations for Head and Neck Radiation Therapy via Graphics Processing Unit–based Dose Deformation Framework

Purpose: The purpose of this study was to systematically monitor anatomic variations and their dosimetric consequences during intensity modulated radiation therapy (IMRT) for head and neck (H&N) cancer by using a graphics processing unit (GPU)-based deformable image registration (DIR) framework. Methods and Materials: Eleven IMRT H&N patients undergoing IMRT with daily megavoltage computed tomography (CT) and weekly kilovoltage CT (kVCT) scans were included in this analysis. Pretreatment kVCTs were automatically registered with their corresponding planning CTs through a GPU-based DIR framework. The deformation of each contoured structure in the H&N region was computed to account for nonrigid change in the patient setup. The Jacobian determinant of the planning target volumes and the surrounding critical structures were used to quantify anatomical volume changes. The actual delivered dose was calculated accounting for the organ deformation. The dose distribution uncertainties due to registration errors were estimated using a landmark-based gamma evaluation. Results: Dramatic interfractional anatomic changes were observed. During the treatment course of 6 to 7 weeks, the parotid gland volumes changed up to 34.7%, and the center-of-mass displacement of the 2 parotid glands varied in the range of 0.9 to 8.8 mm. For the primary treatment volume, the cumulative minimum and mean and equivalent uniform doses assessed by the weekly kVCTs were lower than the planned doses by up to 14.9% (P=.14), 2% (P=.39), and 7.3% (P=.05), respectively. The cumulative mean doses were significantly higher than the planned dose for the left parotid (P=.03) and right parotid glands (P=.006). The computation including DIR and dose accumulation was ultrafast (∼45 seconds) with registration accuracy at the subvoxel level. Conclusions: A systematic analysis of anatomic variations in the H&N region and their dosimetric consequences is critical in improving treatment efficacy. Nearly real-time assessment of anatomic and dosimetric variations is feasible using the GPU-based DIR framework. Clinical implementation of this technology may enable timely plan adaptation and improved outcome.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Evaluation of Historic Nuclear Cloud Measurement for Entrainment Parameter Determination

We evaluated the uncertainties associated with the entrainment parameters computed from nuclear cloud measurements taken during historic weapon testing campaigns. This evaluation found that variations in the way cloud heights and diameters were defined during the analysis of film records result in very large uncertainties in the entrainment parameters, severely limiting the utility of these data. For the measurements of atmospheric conditions at the time of the tests, the main sources of error are associated with the differences in the location and timing of the soundings.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

Analysis of Tasks in Autonomous Systems Using the EMRALD Dynamic Risk Assessment Tool

An autonomous system refers to the system that has the power and ability for self-governance in the performance of system functions. Autonomous systems have been actively pursued in a variety of domains such as automotive, aviation, maritime, medicine, and nuclear fields. As an unmanned concept employing the highest automation level, the autonomous system basically performs most of the work in normal operations or emergency situations. However, despite advances in technology, many researchers have noted these systems still require human actions. The nature of human actions on autonomous systems is different than the human actions that are considered in existing systems. Nevertheless, only a few studies have been conducted on 1) characterizing the different types of errors and risks associated with human actions interacting with autonomous systems and 2) how to evaluate human actions in the autonomous operations. As a starting point, this study aims to investigate differences of tasks in autonomous operation compared to those in existing nuclear power plant operation using the Event Modeling Risk Assessment Using Linked Diagram (EMRALD) software. In this paper, insights aspect of human error and time are derived out and discussed based on the output of the EMRALD models.

99 GENERAL AND MISCELLANEOUS↗

Satellite communication performance evaluation: Computational techniques based on moments

Computational techniques that efficiently compute bit error probabilities when only moments of the various interference random variables are available are presented. The approach taken is a generalization of the well known Gauss-Quadrature rules used for numerically evaluating single or multiple integrals. In what follows, basic algorithms are developed. Some of its properties and generalizations are shown and its many potential applications are described. Some typical interference scenarios for which the results are particularly applicable include: intentional jamming, adjacent and cochannel interferences; radar pulses (RFI); multipath; and intersymbol interference. While the examples presented stress evaluation of bit error probilities in uncoded digital communication systems, the moment techniques can also be applied to the evaluation of other parameters, such as computational cutoff rate under both normal and mismatched receiver cases in coded systems. Another important application is the determination of the probability distributions of the output of a discrete time dynamical system. This type of model occurs widely in control systems, queueing systems, and synchronization systems (e.g., discrete phase locked loops).

Omura, J. K.↗

An Explicit Linear Filtering Solution for the Optimization of Guidance Systems with Statistical Inputs

The determination of optimum filtering characteristics for guidance system design is generally a tedious process which cannot usually be carried out in general terms. In this report a simple explicit solution is given which is applicable to many different types of problems. It is shown to be applicable to problems which involve optimization of constant-coefficient guidance systems and time-varying homing type systems for several stationary and nonstationary inputs. The solution is also applicable to off-design performance, that is, the evaluation of system performance for inputs for which the system was not specifically optimized. The solution is given in generalized form in terms of the minimum theoretical error, the optimum transfer functions, and the optimum transient response. The effects of input signal, contaminating noise, and limitations on the response are included. From the results given, it is possible in an interception problem, for example, to rapidly assess the effects on minimum theoretical error of such factors as target noise and missile acceleration. It is also possible to answer important questions regarding the effect of type of target maneuver on optimum performance.

Stewart, Elwood C.↗

Cognitive Functioning in Space Exploration Missions: A Human Requirement

Solving cognitive issues in the exploration missions will require implementing results from both Human Behavior and Performance, and Space Human Factors Engineering. Operational and research cognitive requirements need to reflect a coordinated management approach with appropriate oversight and guidance from NASA headquarters. First, this paper will discuss one proposed management method that would combine the resources of Space Medicine and Space Human Factors Engineering at JSC, other NASA agencies, the National Space Biomedical Research Institute, Wyle Labs, and other academic or industrial partners. The proposed management is based on a Human Centered Design that advocates full acceptance of the human as a system equal to other systems. Like other systems, the human is a system with many subsystems, each of which has strengths and limitations. Second, this paper will suggest ways to inform exploration policy about what is needed for optimal cognitive functioning of the astronaut crew, as well as requirements to ensure necessary assessment and intervention strategies for the human system if human limitations are reached. Assessment strategies will include clinical evaluation and fitness-to-perform evaluations. Clinical intervention tools and procedures will be available to the astronaut and space flight physician. Cognitive performance will be supported through systematic function allocation, task design, training, and scheduling. Human factors requirements and guidelines will lead to well-designed information displays and retrieval systems that reduce crew time and errors. Means of capturing process, design, and operational requirements to ensure crew performance will be discussed. Third, this paper will describe the current plan of action, and future challenges to be resolved before a lunar or Mars expedition. The presentation will include a proposed management plan for research, involvement of various organizations, and a timetable of deliverables.

Fiedler, Edan↗

Using Relative Position and Temporal Judgments to Assess the Effects of Texture and Field of View on Spatial Awareness for Synthetic Vision Systems Displays

Synthetic Vision Systems (SVS) depict computer generated views of terrain surrounding an aircraft. In the assessment of textures and field of view (FOV) for SVS, no studies have directly measured the 3 levels of spatial awareness: identification of terrain, its relative spatial location, and its relative temporal location. This work introduced spatial awareness measures and used them to evaluate texture and FOV in SVS displays. Eighteen pilots made 4 judgments (relative angle, distance, height, and abeam time) regarding the location of terrain points displayed in 112 5-second, non-interactive simulations of a SVS heads down display. Texture produced significant main effects and trends for the magnitude of error in the relative distance, angle, and abeam time judgments. FOV was significant for the directional magnitude of error in the relative distance, angle, and height judgments. Pilots also provided subjective terrain awareness ratings that were compared with the judgment based measures. The study found that elevation fishnet, photo fishnet, and photo elevation fishnet textures best supported spatial awareness for both the judgments and the subjective awareness measures.

Bolton, Matthew L.↗

Time Accurate Unsteady Pressure Loads Simulated for the Space Launch System at a Wind Tunnel Condition

Using the Fully Unstructured Three-Dimensional (FUN3D) computational fluid dynamics code, an unsteady, time-accurate flow field about a Space Launch System configuration was simulated at a transonic wind tunnel condition (Mach = 0.9). Delayed detached eddy simulation combined with Reynolds Averaged Naiver-Stokes and a Spallart-Almaras turbulence model were employed for the simulation. Second order accurate time evolution scheme was used to simulate the flow field, with a minimum of 0.2 seconds of simulated time to as much as 1.4 seconds. Data was collected at 480 pressure taps at locations, 139 of which matched a 3% wind tunnel model, tested in the Transonic Dynamic Tunnel (TDT) facility at NASA Langley Research Center. Comparisons between computation and experiment showed agreement within 5% in terms of location for peak RMS levels, and 20% for frequency and magnitude of power spectral densities. Grid resolution and time step sensitivity studies were performed to identify methods for improved accuracy comparisons to wind tunnel data. With limited computational resources, accurate trends for reduced vibratory loads on the vehicle were observed. Exploratory methods such as determining minimized computed errors based on CFL number and sub-iterations, as well as evaluating frequency content of the unsteady pressures and evaluation of oscillatory shock structures were used in this study to enhance computational efficiency and solution accuracy. These techniques enabled development of a set of best practices, for the evaluation of future flight vehicle designs in terms of vibratory loads.

Alter, Stephen J.↗

Efficient surrogate modeling methods for large-scale Earth system models based on machine-learning techniques

Abstract. Improving predictive understanding of Earth system variability and change requires data–model integration. Efficient data–model integration for complex models requires surrogate modeling to reduce model evaluation time. However, building a surrogate of a large-scale Earth system model (ESM) with many output variables is computationally intensive because it involves a large number of expensive ESM simulations. In this effort, we propose an efficient surrogate method capable of using a few ESM runs to build an accurate and fast-to-evaluate surrogate system of model outputs over large spatial and temporal domains. We first use singular value decomposition to reduce the output dimensions and then use Bayesian optimization techniques to generate an accurate neural network surrogate model based on limited ESM simulation samples. Our machine-learning-based surrogate methods can build and evaluate a large surrogate system of many variables quickly. Thus, whenever the quantities of interest change, such as a different objective function, a new site, and a longer simulation time, we can simply extract the information of interest from the surrogate system without rebuilding new surrogates, which significantly reduces computational efforts. We apply the proposed method to a regional ecosystem model to approximate the relationship between eight model parameters and 42 660 carbon flux outputs. Results indicate that using only 20 model simulations, we can build an accurate surrogate system of the 42 660 variables, wherein the consistency between the surrogate prediction and actual model simulation is 0.93 and the mean squared error is 0.02. This highly accurate and fast-to-evaluate surrogate system will greatly enhance the computational efficiency of data–model integration to improve predictions and advance our understanding of the Earth system.

58 GEOSCIENCES↗

Stochastic process approximation for recursive estimation with guaranteed bound on the error covariance

An approach, is proposed for the design of approximate, fixed order, discrete time realizations of stochastic processes from the output covariance over a finite time interval, was proposed. No restrictive assumptions are imposed on the process; it can be nonstationary and lead to a high dimension realization. Classes of fixed order models are defined, having the joint covariance matrix of the combined vector of the outputs in the interval of definition greater or equal than the process covariance; (the difference matrix is nonnegative definite). The design is achieved by minimizing, in one of those classes, a measure of the approximation between the model and the process evaluated by the trace of the difference of the respective covariance matrices. Models belonging to these classes have the notable property that, under the same measurement system and estimator structure, the output estimation error covariance matrix computed on the model is an upper bound of the corresponding covariance on the real process. An application of the approach is illustrated by the modeling of random meteorological wind profiles from the statistical analysis of historical data.

Menga, G.↗

Combining Sparse Approximate Factorizations with Mixed-precision Iterative Refinement

The standard LU factorization-based solution process for linear systems can be enhanced in speed or accuracy by employing mixed-precision iterative refinement. Most recent work has focused on dense systems. We investigate the potential of mixed-precision iterative refinement to enhance methods for sparse systems based on approximate sparse factorizations. In doing so, we first develop a new error analysis for LU- and GMRES-based iterative refinement under a general model of LU factorization that accounts for the approximation methods typically used by modern sparse solvers, such as low-rank approximations or relaxed pivoting strategies. We then provide a detailed performance analysis of both the execution time and memory consumption of different algorithms, based on a selected set of iterative refinement variants and approximate sparse factorizations. Our performance study uses the multifrontal solver MUMPS, which can exploit block low-rank factorization and static pivoting. We evaluate the performance of the algorithms on large, sparse problems coming from a variety of real-life and industrial applications showing that mixed-precision iterative refinement combined with approximate sparse factorization can lead to considerable reductions of both the time and memory consumption.

97 MATHEMATICS AND COMPUTING↗