Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “load monitoring”

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 811 records · Page 45
All sourcesNASA NTRSDOE OSTI
NASA NTRS1985

Characterization of Mode I and mixed-mode delamination growth in T300/5208 graphite/epoxy

The roles played by Mode I and Mode II strain-energy release rates (G-I and G-II, respectively) in inducing delamination growth under static and fatigue loading were investigated, using T300/5208 graphite/epoxy specimens. Double cantilever beam (DCB) specimens and cracked lap shear (CLS) specimens were used for pure Mode I and mixed-mode tests, respectively. Fatigue-induced delamination growth was characterized by constant-amplitude fatigue tests at a minimum to maximum cyclic load ratio of 0.05 and a frequency of 10 Hz. During the tests, the maximum and minimum strain-energy release rates (Gmax, Gmin) and the delamination growth rate (da/dN) were monitored. Static tests on mixed-mode CLS specimens measured the total strain-energy release rate, which was broken into G-I and G-II components using finite-element analysis. A power-law relationship between da/dN and G-Imax, and da/dN and Gmax were obtained from fatigue test results on DCB and CLS specimens, respectively. The power law for a pure Mode II delamination was derived from CLS results by subtracting the contribution due to G-I.

Ramkumar, R. L.↗
NASA NTRS2000

Thermoelastic Stress Analysis: An NDE Tool for the Residual Stress Assessment of Metallic Alloys

During manufacturing, certain propulsion components that will be used in a cyclic fatigue environment are fabricated to contain compressive residual stresses on their surfaces because these stresses inhibit the nucleation of cracks. Overloads and elevated temperature excursions cause the induced residual stresses to dissipate while the component is still in service, lowering its resistance to crack initiation. Research at the NASA Glenn Research Center at Lewis Field has focused on employing the Thermoelastic Stress Analysis technique (TSA, also recognized as SPATE: Stress Pattern Analysis by Thermal Emission) as a tool for monitoring the residual stress state of propulsion components. TSA is based on the fact that materials experience small temperature changes when they are compressed or expanded. When a structure is cyclically loaded (i.e., cyclically compressed and expanded), the resulting surface-temperature profile correlates to the stress state of the structure s surface. The surface-temperature variations resulting from a cyclic load are measured with an infrared camera. Traditionally, the temperature amplitude of a TSA signal has been theoretically defined to be linearly dependent on the cyclic stress amplitude. As a result, the temperature amplitude resulting from an applied cyclic stress was assumed to be independent of the cyclic mean stress.

Gyekenyesi, Andrew L.↗
NASA NTRS2013

Advanced Neutron Spectrometer

Energetic neutron measurements remain a challenge for space science investigations and radiation monitoring for human exploration beyond LEO. We are investigating a new composite scintillator design that uses Li6 glass scintillator embedded in a PVT block. A comparison between Li6 and Boron 10 loaded scintillators are being studied to assess the advantages and shortcomings of these two techniques. We present the details of the new Li6 design and results from the comparison of the B10 and Li6 techniques during exposures in a mixed radiation field produced by high energy protons interacting in a target material.

Christl, Mark↗
NASA NTRS2015

Designing an Exploration Atmosphere Prebreathe Protocol

Extravehicular activities (EVAs) at remote locations must maximize limited resources such as oxygen (O2) and also minimize the risk of decompression sickness (DCS). A proposed remote denitrogenation (prebreathe) protocol requires astronauts to live in a mildly hypoxic atmosphere at 8.2 psia while periodically performing EVAs at 4.3 psia. Empirical data are required to confirm that the protocol meets the current accept requirements: less than or equal to 15% incidence of Type I DCS, less than or equal to 20% incidence of Grade IV venous gas emboli (VGE), both at 95% statistical confidence, with no Type II DCS symptom during the validation trial. METHODS: A repeated measures statistical design is proposed in which groups of 6 subjects with physical characteristics similar to active-duty astronauts would first become equilibrated to an 8.2 psia atmosphere in a hypobaric chamber containing 34% O2 and 66% N2, over 48 h, and then perform 4 simulated EVAs at 4.3 psia over the next 9 days. In the equilibration phase, subjects undergo a 3-h 100% O2 mask prebreathe prior to and during a 5-min ascent to 8.2 psia to prevent significant tissue N2 supersaturation on reaching 8.2 psia. Masks would be removed once 34% O2 is established at 8.2 psia, and subjects would then equilibrate to this atmosphere for 48 h. The hypoxia is equivalent to breathing air at 1,220 meters (4,000 ft) altitude, just as was experienced in the shuttle 10.2 psia - 26.5% O2 staged denitrogenation protocol and the current ISS campout denitrogenation protocol. For simulated EVAs, each subject dons a mask and breathes 85% O2 and 15% N2 during a 3-min depressurization to 6.0 psia, holds for 15 min, and then completes a 3-min depressurization to 4.3 psia. The simulated EVA period starts when 6.0 psia is reached and continues for a total of 240 min (222 min at 4.3 psia). During this time, subjects will follow a prescribed repetitive activity against loads in the upper and lower body with mean metabolic rate approaching 1500 BTU/hr [378 kcal/hr (O2 consumption about 1.3 l(sub STPD)/min)] in ambulatory subjects. Noninvasive Doppler ultrasound bubble monitoring for VGE in the pulmonary artery will be performed on subjects by 2 Doppler Technicians at about 15 min intervals while at 4.3 psia. At the end of this period, a 15-min repressurization returns all subjects back to 8.2 psia and the cycle is repeated 3 additional times with a day of rest between simulated EVAs. RESULTS: With an assumed 1.5% probability of DCS [P(DCS)] and accounting for within-subject correlation, running the proposed study with 20 subjects has a 95% probability of meeting the accept criterion for DCS. But if the true probability of DCS is 3.0%, then 30 subjects would be needed to achieve about the same probability to meet our accept criterion. These results assume a standard deviation of 1.4 for the between-subjects random component of P(DCS) on a logit scale, which was estimated from a previous study.

Conkin, Johnny↗
NASA NTRS2024

The Diurnal Variation of Dust and Water Ice Aerosol Optical Depth at Jezero Crater Observed By MEDA/TIRS Over A Full Martian Year

The Thermal InfraRed Sensor (TIRS) on the Perseverance rover has provided nearly two full Mars years of systematic monitoring of the total aerosol optical depth above Jezero Crater. These observations span a wide range of timescales, capturing seasonal patterns, diurnal variations, and minute-to-minute fluctuations in aerosol loading. By combining TIRS retrievals with orbital observations, the relative contributions of dust and water ice aerosols can be estimated, revealing their different seasonal and diurnal behaviors. The TIRS record shows distinct periods of dust storm activity, including strong regional storms during the perihelion season as well as short lived but intense dust events outside the typical dust storm season. Water ice clouds exhibit pronounced seasonal and diurnal variability, with peak activity occurring during the aphelion season but with a presence throughout the year. The diurnal variation of clouds differs significantly between the aphelion and perihelion seasons, with clouds persisting throughout the night during the aphelion season, while largely absent outside of specific periods after sunrise and sunset during the perihelion season. These results provide new insights into the complex behavior of aerosols at Jezero Crater and their connections to atmospheric dynamics and the Martian dust and water cycles.

Michael D Smith↗
DOE OSTI2020

Greggd

greg(g)d - Global runtime for eBPF-enabled gathering (w/ gumption) daemon Recently the linux kernel has added support for low-level kernel monitoring and profiling through a in-kernel virtual machine. The tooling around these new features (the extended Berkley Packet Filter or eBPF for short) is not mature and is difficult to use. Benefits from eBPF are especially hard to realize while trying to do large scale deployments and integrate with existing metric analysis stacks. A tool was needed to enable loading and collecting data from eBPF programs on large scale HPC systems. Given the problems above it was obvious we needed some wrapper program to compile, load, and collect data from eBPF programs running in the kernel. This tool needed to be lightweight without a heavy set of dependencies, relatively stable between different kernel versions, and integrate nicely with existing widely used metric collection tools. We wrote a program that wraps the eBPF tooling and sends data to our metric gathering tool. eBPF programs are either compiled using the host compiler stack, or loaded in the kernel directly from an object file. These programs are then attached to the system calls that we want to profile. Whenever these system calls are run, the eBPF program collects information of interest and writes that to memory. Our wrapper program polls these memory locations, reads and formats the data, then sends the information to a local unix socket. Our other monitoring tools are configured to read from that socket and send it to the rest of our metric monitoring stack for analysis.

Voss, Joseph [Oak Ridge National Lab. (ORNL), Oak ↗
NASA NTRSDate not supplied

Trustworthy Machine Learning for Damage Identification in Composites

A challenging opportunity in structural health monitoring of composite materials is using machine learning (ML) methods to classify acoustic emissions (AE) according to the damage mechanism that emitted the signal. Although a wide variety of ML frameworks have been developed, there is a distinct lack of ground truth datasets which has precluded any direct assessment of their accuracy. Here, we present a novel ground truth dataset gathered on simplified unidirectional SiC/SiC composite structures. Herein, AE is collected from minicomposites which are loaded to targeted percentages of the ultimate tensile stress. These minicomposites are then volumetrically imaged with XCT and individual damage events, along with the mechanism, are correlated to AE. We explore the signal features that allow for mechanism discrimination, along with the feasibility of both unsupervised and supervised frameworks for use in the online monitoring of composite structures.

Machine learning, acoustic emission, ceramic matri↗
NASA NTRS2004

Characterization of microbial and chemical composition of shuttle wet waste with permanent gas and volatile organic compound analyses

Solid-waste treatment in space for Advanced Life Support, ALS, applications requires that the material can be safely processed and stored in a confined environment. Many solid-wastes are not stable because they are wet (40-90% moisture) and contain levels of soluble organic compounds that can contribute to the growth of undesirable microorganisms with concomitant production of noxious odors. In the absence of integrated Advanced Life Support systems on orbit, permanent gas, trace volatile organic and microbiological analyses were performed on crew refuse returned from the volume F "wet" trash of three consecutive Shuttle missions (STS-105, 109, and 110). These analyses were designed to characterize the short-term biological stability of the material and assess potential crew risks resulting from microbial decay processes during storage. Waste samples were collected post-orbiter landing and sorted into packaging material, food waste, toilet waste, and bulk liquid fractions deposited during flight in the volume F container. Aerobic and anaerobic microbial loads were determined in each fraction by cultivation on R2A and by acridine orange direct count (AODC). Dry and ash weights were performed to determine both water and organic content of the materials. Experiments to determine the aerobic and anaerobic biostability of refuse stored for varying periods of time were performed by on-line monitoring of CO2 and laboratory analysis for production of hydrogen sulfide and methane. Volatile organic compounds and permanent gases were analyzed using EPA Method TO15 by USEPA et al. [EPA Method TO15, The Determination of Volatile Organic Compounds (VOCs) in Ambient Air using SUMMA, Passivated Canister Sampling and Gas Chromatographic Analysis,1999] with gas chromatography/mass spectrometry and by gas chromatography with selective detectors. These baseline measures of waste stream content, labile organics, and microbial load in the volume F Shuttle trash provide data for waste subsystem analysis and atmospheric management within the ALS Project. Published by Elsevier Ltd on behalf of COSPAR.

NASA Center KSC↗
DOE OSTI2023

Sensitivity measurements for a 250 MHz quartz shear-horizontal surface acoustic wave biosensor under liquid viscous loading

Surface acoustic wave (SAW) devices have been used in biochemical assays due to their high sensitivity. The device sensitivity is a function of changes in the density and viscosity of the liquid. Here, we studied the effect of fluid viscosity using a 250 MHz quartz shear-horizontal (SH)-SAW biosensor by monitoring different concentrations of binary aqueous/glycerol solutions. In this study, the sensitivity of the biosensor was determined by fitting the data to models derived from perturbation theory. Measurements in water were used as the reference. For a 0% to 50% glycerol solution, an 87°–204° separation in the phase shift was observed. The slope of the plot of the phase shift vs (ηρ)0.5 was used to indicate the sensor’s sensitivity. The sensitivity for our 250 MHz quartz SH-SAW sensors was calculated to be 3.7×10−3m2sKg. The corresponding mass sensitivity was determined to be 9.25 × 105m2Kg. The limit of detection was calculated to be 36 picograms (pg), while the limit of quantification or LOQ was calculated to be 109 pg. Traditionally, liquid phase measurements have been challenging for SAW devices because liquids dampen the vibrating sensors severely. This problem has been largely solved using a transverse (shear) wave instead of the more popular longitudinal or Rayleigh waves. Liquid measurements are now possible using transverse waves, also known as shear waves, because transverse waves are only minimally attenuated by liquids. Shear-horizontal SAW sensors (SH-SAW) show great promise as label-free biosensors because of their ability to handle liquid samples. However, the viscosity of the liquid still induces loading effects and can be measured when the liquid is loaded onto the SH-SAW propagating surface (delay line). When the liquid above the delay line is perturbed by physical or chemical changes, such as binding to a receptor, it alters the propagating acoustic wave. The SH-SAW device can measure these changes in liquid properties as a change in the wave’s phase compared to the original wave. The device’s phase shift was recorded as a function of the changes in the density and viscosity of the binary glycerol solution and used to determine the sensitivity in the linear dynamic range of responses.

Materials Science↗
DOE OSTI2020

Performance of a CO 2 sorbent for indoor air cleaning applications: Effects of environmental conditions, sorbent aging, and adsorption of co-occurring formaldehyde

Indoor air cleaning systems that incorporate CO 2 sorbent materials enable HVAC load shifting and efficiency improvements. This study developed a bench-scale experimental system to evaluate the performance of a sorbent under controlled operation conditions. A thermostatic holder containing 3.15 g sorbent was connected to a manifold that delivered CO 2 -enriched air at a known temperature and relative humidity (RH). The air stream was also enriched with 0.8-2.1 ppm formaldehyde. The CO 2 concentration was monitored in real-time upstream and downstream of the sorbent, and integrated formaldehyde samples were collected at different times using DNPH-coated silica cartridges. Sorbent regeneration was carried out by circulating clean air in countercurrent. Almost 200 loading/regeneration cycles were performed in the span of 17 months, from which 104 were carried out at reference test conditions defined by loading with air at 25°C, 38% RH, and 1000 ppm CO 2 , and regenerating with air at 80°C, 3% RH and 400 ppm CO 2 . The working capacity decreased slightly from 43-44 mg CO 2 per g sorbent to 39-40 mg per g over the 17 months. The capacity increased with lower loading temperature (in the range 15-35°C) and higher regeneration temperature, between 40 and 80°C. The CO 2 capacity was not sensitive to the moisture content in the range 6-9 g/m 3 , and decreased slightly when dry air was used. Loading isothermal breakthrough curves were fitted to three simple adsorption models, verifying that pseudo-first-order kinetics appropriately describes the adsorption process. The model predicted that equilibrium capacities decreased with increasing temperature from 15 to 35°C, while adsorption rate constants slightly increased. The formaldehyde adsorption efficiency was 80%-99% in different cycles, corresponding to an average capacity of 86 ± 36 µg/g. Formaldehyde was not quantitatively released during regeneration, but its accumulation on the sorbent did not affect CO 2 adsorption.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Coupled Thermo-Hydrological-Mechanical-Chemical Behavior of Anisotropic Granite for Geologic Disposal of High-Level Radioactive Waste: A Core-Scale Laboratory Investigation

The coupled thermo-hydrological-mechanical-chemical (THMC) behavior of rock within an Excavation Damaged Zone (EDZ) is critical for the safety and long-term performance of a geological repository for high-level radioactive wastes. While many laboratory experiments have been conducted to investigate EDZ rocks, the flow and deformation characteristics resulting from anisotropic rock textures and microcrack distribution under triaxial loading and elevated temperatures remain poorly understood. Particularly, cracks at various scales serve as fast paths for fluid flow and solute transport and present as focal points of mechanical weakness, which complicate the coupled THMC processes in anisotropic EDZ rocks and challenge modeling predictions. Here, in this study, a series of core-scale experiments was conducted on three granite samples under repository-relevant conditions. These rock samples were obtained from the Grimsel Underground Research Laboratory (URL), featured by anisotropic minerals (represented by bedding layers) and microcrack distributions and coarse cm-scale grain sizes. During the experiments, samples were subjected to an elevated temperature at 90 °C and different triaxial loading conditions either by radial (normal to bedding layers) or axial (parallel to bedding layers) compaction. Water was injected into the samples, and the rock permeability evolutions and effluent water chemistry were monitored closely. For intact samples, thermal expansion of minerals at 90 °C resulted in a large, 75% irreversible permeability reduction and rock strengthening under radial compaction, while thermal impact was limited to a 15% permeability reduction under axial compaction. In contrast, for a sample containing open cracks, the growth of fractures during the experiment resulted in an abrupt permeability increase and fast failure at 90 °C. The effluent water chemistry indicates much more considerable mineral dissolution from large shear sliding than that in rocks dominated by mechanical compaction. These results helped better understand the coupled THMC processes in anisotropic rocks containing cracks, evaluate the behaviors of EDZ rocks, and predict the long-term evolution of EDZ for the performance of the repositories.

Coupled THMC processes↗
NASA NTRS1975

Tug and payload-to-Orbiter interface requirements

This paper presents results of a study to identify, evaluate, and develop Tug plus payload-to-Orbiter accommodation requirements. Interface areas include structural, mechanical, fluids, environmental, avionics, and safety. Each interface area was investigated to determine the best physical and operational method of accomplishing the required functions with an overriding goal of establishing sample and flexible interface requirements suitable for Tug, Tug payloads, and other cargo. The recommended system for supporting the Tug and its payload from the Orbiter employs a cylindrical load-carrying structure called a deployment adapter that has a rotational capability for Tug deployment. The design and operation of the deployment adapter and other peripheral interface equipment such as crew compartment monitor and control panels and cargo bay fluid and electrical interface panels are discussed.

Bock, E. H.↗
NASA NTRS1988

Cryogenic systems for the large deployable reflector

There are five technologies which may have application for Large Deployable Reflector (LDR), one passive and four active. In order of maturity, they are passive stored cryogen systems, and mechanical, sorption, magnetic, and pulse-tube refrigerators. In addition, deep space radiators will be required to reject the heat of the active systems, and may be useful as auxiliary coolers for the stored cryogen systems. Hybrid combinations of these technologies may well be more efficient than any one alone, and extensive system studies will be required to determine the best trade-offs. Stored cryogen systems were flown on a number of missions. The systems are capable of meeting the temperature requirements of LDR. The size and weight of stored cryogen systems are proportional to heat load and, as a result, are applicable only if the low-temperature heat load can be kept small. Systems using chemisorption and physical adsorption for compressors and pumps have received considerable attention in the past few years. Systems based on adiabatic demagnetization of paramagnetic salts were used for refrigeration for many years. Pulse-tube refrigerators were recently proposed which show relatively high efficiency for temperatures in the 60 to 80 K range. The instrument heat loads and operating temperatures are critical to the selection and design of the cryogenic system. Every effort should be made to minimize heat loads, raise operating temperatures, and to define these precisely. No one technology is now ready for application to LDR. Substantial development efforts are underway in all of the technologies and should be monitored and advocated. Magnetic and pulse-tube refrigerators have high potential.

Mason, Peter V.↗
NASA NTRS1993

Vortex flap flight test operations, a safe approach

A flight test experiment was conducted at the Langley Research Center to evaluate a wing leading-edge vortex flap concept designed for use on an aircraft with highly swept wings. The flap concept was designed as a modification to the wing leading edge of an F-106B airplane. The flight testing required operations at conditions that would exceed the structural load envelope of the basic airplane in order to acquire desired research data for the modified configuration. Accordingly, the operational envelope of the modified aircraft was incrementally expanded and real-time monitoring of airframe strains at critical wing locations was mandated to insure safety of flight. The flight tests were conducted in two phases: Phase I to establish baseline data with the unmodified wing, and Phase II to determine the effects of the vortex flap on performance, handling qualities, and flow field characteristics. This paper focuses on a description of the approach and procedures used to provide the strain-gage monitoring to insure structural integrity. Highlights of the wing modification and the overall operation are also included. Within a -year period, 110 research flights were successfully completed, providing researchers with sufficient data to assess the potential benefits ascribed to the vortex flap concept without encountering severe structural problems or mishaps.

Dicarlo, Daniel J.↗
NASA NTRS1995

Improved Speed Control System for the 87,000 HP Wind Tunnel Drive

This paper describes the design, installation, and integrated systems tests for a new drive motor speed control system which was part of a recent rehab project for the NASA Lewis 8x6 Supersonic Wind Tunnel. The tunnel drive consists of three mechanically-coupled 29,000 HP wound rotor induction motors driving an axial flow compressor. Liquid rheostats are used to vary the impedance of the rotor circuits, thus varying the speed of the drive system. The new design utilizes a distributed digital control system with a dual touch screen CRT operator console to provide alarm monitoring, logging, and trending. The liquid rheostats are driven by brushtype servomotor systems with magnetostrictive linear displacement transducers used for position feedback. The new system achieved all goals for speed variations with load, motor load balance, and control of total power.

Becks, Edward A.↗
NASA NTRS1992

Autonomous power system brassboard

The Autonomous Power System (APS) brassboard is a 20 kHz power distribution system which has been developed at NASA Lewis Research Center, Cleveland, Ohio. The brassboard exists to provide a realistic hardware platform capable of testing artificially intelligent (AI) software. The brassboard's power circuit topology is based upon a Power Distribution Control Unit (PDCU), which is a subset of an advanced development 20 kHz electrical power system (EPS) testbed, originally designed for Space Station Freedom (SSF). The APS program is designed to demonstrate the application of intelligent software as a fault detection, isolation, and recovery methodology for space power systems. This report discusses both the hardware and software elements used to construct the present configuration of the brassboard. The brassboard power components are described. These include the solid-state switches (herein referred to as switchgear), transformers, sources, and loads. Closely linked to this power portion of the brassboard is the first level of embedded control. Hardware used to implement this control and its associated software is discussed. An Ada software program, developed by Lewis Research Center's Space Station Freedom Directorate for their 20 kHz testbed, is used to control the brassboard's switchgear, as well as monitor key brassboard parameters through sensors located within these switches. The Ada code is downloaded from a PC/AT, and is resident within the 8086 microprocessor-based embedded controllers. The PC/AT is also used for smart terminal emulation, capable of controlling the switchgear as well as displaying data from them. Intelligent control is provided through use of a T1 Explorer and the Autonomous Power Expert (APEX) LISP software. Real-time load scheduling is implemented through use of a 'C' program-based scheduling engine. The methods of communication between these computers and the brassboard are explored. In order to evaluate the features of both the brassboard hardware and intelligent controlling software, fault circuits have been developed and integrated as part of the brassboard. A description of these fault circuits and their function is included. The brassboard has become an extremely useful test facility, promoting artificial intelligence (AI) applications for power distribution systems. However, there are elements of the brassboard which could be enhanced, thus improving system performance. Modifications and enhancements to improve the brassboard's operation are discussed.

Merolla, Anthony↗
NASA NTRS1975

Atmospheric particulate measurements in Norfolk, Virginia

Characterization of atmospheric particulates was conducted at a site near the center of Norfolk, Virginia. Air quality was measured in terms of atmospheric mass loading, particle size distribution, and particulate elemental composition for a period of 2 weeks. The objectives of this study were (1) to establish a mean level of air quality and deviations about this mean, (2) to ascertain diurnal changes or special events in air quality, and (3) to evaluate instrumentation and sampling schedules. Simultaneous measurements were made with the following instruments: a quartz crystal microbalance particulate monitor, a light-scattering multirange particle counter, a high-volume air sampler, and polycarbonate membrane filters. To assess the impact of meteorological conditions on air quality variations, continuous data on temperature, relative humidity, wind speed, and wind direction were recorded. Particulate elemental composition was obtained from neutron activation and scanning electron microscopy analyses of polycarbonate membrane filter samples. The measured average mass loading agrees reasonably well with the mass loadings determined by the Virginia State Air Pollution Control Board. There are consistent diurnal increases in atmospheric mass loading in the early morning and a sample time resolution of 1/2 hour seems necessary to detect most of the significant events.

Storey, R. W., Jr.↗
DOE OSTI2023

Reinforcement Learning for Load-balanced Parallel Particle Tracing

We explore an online reinforcement learning (RL) paradigm to dynamically optimize parallel particle tracing performance in distributed-memory systems. Our method combines three novel components: (1) a work donation algorithm, (2) a high-order workload estimation model, and (3) a communication cost model. First, we design an RL-based work donation algorithm. Our algorithm monitors workloads of processes and creates RL agents to donate data blocks and particles from high-workload processes to low-workload processes to minimize program execution time. The agents learn the donation strategy on the fly based on reward and cost functions designed to consider processes' workload changes and data transfer costs of donation actions. Second, we propose a workload estimation model, helping RL agents estimate the workload distribution of processes in future computations. Third, we design a communication cost model that considers both block and particle data exchange costs, helping RL agents make effective decisions with minimized communication costs. We demonstrate that our algorithm adapts to different flow behaviors in large-scale fluid dynamics, ocean, and weather simulation data. Our algorithm improves parallel particle tracing performance in terms of parallel efficiency, load balance, and costs of I/O and communication for evaluations with up to 16,384 processors.

Distributed and parallel particle tracing↗