Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Real--Time Dynamic Modeling”

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 523 records · Page 29
All sourcesNASA NTRSDOE OSTI
NASA NTRS1993

Real-time dynamic simulation of the Cassini spacecraft using DARTS. Part 2: Parallel/vectorized real-time implementation

Part 1 of this paper presented the requirements for the real-time simulation of Cassini spacecraft along with some discussion of the DARTS algorithm. Here, in Part 2 we discuss the development and implementation of parallel/vectorized DARTS algorithm and architecture for real-time simulation. Development of the fast algorithms and architecture for real-time hardware-in-the-loop simulation of spacecraft dynamics is motivated by the fact that it represents a hard real-time problem, in the sense that the correctness of the simulation depends on both the numerical accuracy and the exact timing of the computation. For a given model fidelity, the computation should be computed within a predefined time period. Further reduction in computation time allows increasing the fidelity of the model (i.e., inclusion of more flexible modes) and the integration routine.

Fijany, A.↗
NASA NTRS1993

Streak camera based SLR receiver for two color atmospheric measurements

To realize accurate two-color differential measurements, an image digitizing system with variable spatial resolution was designed, built, and integrated to a photon-counting picosecond streak camera, yielding a temporal scan resolution better than 300 femtosecond/pixel. The streak camera is configured to operate with 3 spatial channels; two of these support green (532 nm) and uv (355 nm) while the third accommodates reference pulses (764 nm) for real-time calibration. Critical parameters affecting differential timing accuracy such as pulse width and shape, number of received photons, streak camera/imaging system nonlinearities, dynamic range, and noise characteristics were investigated to optimize the system for accurate differential delay measurements. The streak camera output image consists of three image fields, each field is 1024 pixels along the time axis and 16 pixels across the spatial axis. Each of the image fields may be independently positioned across the spatial axis. Two of the image fields are used for the two wavelengths used in the experiment; the third window measures the temporal separation of a pair of diode laser pulses which verify the streak camera sweep speed for each data frame. The sum of the 16 pixel intensities across each of the 1024 temporal positions for the three data windows is used to extract the three waveforms. The waveform data is processed using an iterative three-point running average filter (10 to 30 iterations are used) to remove high-frequency structure. The pulse pair separations are determined using the half-max and centroid type analysis. Rigorous experimental verification has demonstrated that this simplified process provides the best measurement accuracy. To calibrate the receiver system sweep, two laser pulses with precisely known temporal separation are scanned along the full length of the sweep axis. The experimental measurements are then modeled using polynomial regression to obtain a best fit to the data. Data aggregation using normal point approach has provided accurate data fitting techniques and is found to be much more convenient than using the full rate single shot data. The systematic errors from this model have been found to be less than 3 ps for normal points.

Varghese, Thomas K.↗
NASA NTRS1993

A SIMULINK environment for flight dynamics and control analysis: Application to the DHC-2 Beaver. Part 1: Implementation of a model library in SIMULINK. Part 2: Nonlinear analysis of the Beaver autopilot

The design of advanced Automatic Aircraft Control Systems (AACS's) can be improved upon considerably if the designer can access all models and tools required for control system design and analysis through a graphical user-interface, from within one software environment. This MSc-thesis presents the first step in the development of such an environment, which is currently being done at the Section for Stability and Control of Delft University of Technology, Faculty of Aerospace Engineering. The environment is implemented within the commercially available software package MATLAB/SIMULINK. The report consists of two parts. Part I gives a detailed description of the AACS design environment. The heart of this environment is formed by the SIMULINK implementation of a nonlinear aircraft model in block-diagram format. The model has been worked out for the old laboratory aircraft of the Faculty, the De Havilland DHC-2 'Beaver', but due to its modular structure, it can easily be adapted for other aircraft. Part I also describes MATLAB programs which can be applied for finding steady-state trimmed-flight conditions and for linearization of the aircraft model, and it shows how the built-in simulation routines of SIMULINK have been used for open-loop analysis of the aircraft dynamics. Apart from the implementation of the models and tools, a thorough treatment of the theoretical backgrounds is presented. Part II of this report presents a part of an autopilot design process for the 'Beaver' aircraft, which clearly demonstrates the power and flexibility of the AACS design environment from part I. Evaluations of all longitudinal and lateral control laws by means of nonlinear simulations are treated in detail. The AACS design environment from part I proved to be a very useful tool for designing the control laws of the 'Beaver' autopilot within a very tight time-schedule. The autopilot design process itself will be used as a guideline for future AACS research at the Faculty of Aerospace Engineering. Flight tests of the 'Beaver' autopilot, done after evaluating the control laws in the SIMULINK package, proved to be quite successful. In the future, the AACS design package will evolve into a standardized, integrated design environment which can be applied to virtually any type of aircraft. The AACS design cycle will be shortened further by developing tools for automatically porting control laws from the MATLAB/SIMULINK environment to a piloted real-time flight simulator and the Flight Control Computers of the aircraft.

Flight Control System Design↗
NASA NTRS1983

Transient responses of phosphoric acid fuel cell power plant system

An analytical and computerized study of the steady state and transient response of a phosphoric acid fuel cell (PAFC) system was completed. Parametric studies and sensitivity analyses of the PAFC system's operation were accomplished. Four non-linear dynamic models of the fuel cell stack, reformer, shift converters, and heat exchangers were developed based on nonhomogeneous non-linear partial differential equations, which include the material, component, energy balance, and electrochemical kinetic features. Due to a lack of experimental data for the dynamic response of the components only the steady state results were compared with data from other sources, indicating reasonably good agreement. A steady state simulation of the entire system was developed using, nonlinear ordinary differential equations. The finite difference method and trial-and-error procedures were used to obtain a solution. Using the model, a PAFC system, that was developed under NASA Grant, NCC3-17, was improved through the optimization of the heat exchanger network. Three types of cooling configurations for cell plates were evaluated to obtain the best current density and temperature distributions. The steady state solutions were used as the initial conditions in the dynamic model. The transient response of a simplified PAFC system, which included all of the major components, subjected to a load change was obtained. Due to the length of the computation time for the transient response calculations, analysis on a real-time computer was not possible. A simulation of the real-time calculations was developed on a batch type computer. The transient response characteristics are needed for the optimization of the design and control of the whole PAFC system. All of the models, procedures and simulations were programmed in Fortran and run on IBM 370 computers at Cleveland State University and the NASA Lewis Research Center.

Lu, Cheng-Yi↗
NASA NTRS1996

Two-Dimensional Dynamic Simulation of a Continuous Foil Bearing

In this paper, the two dimensional(radial and circumferential) transient Navier-Stokes equations are used to solve the hydrodynamic problem in conjunction with the time dependent motion of the journal, and the deformable, spring supported foil. The elastic deformation of the foil and its supports are simulated by a finite element model. The time-dependent Navier-Stokes formulation is used to solve for the interaction between the fluid lubricant, the motion of the journal and the deformable foil boundary. The steady state, the quasi-transient and the full transient dynamic simulation of the foil-fluid journal interaction are examined on a comparative basis. For the steady state simulation, the fluid lubricant pressures are evaluated for a particular journal position, by means of an iterative scheme until convergence is achieved in both the fluid pressures and the corresponding foil deformation. For the quasi-transient case, the transient motion of the journal is calculated using a numerical integration scheme for the velocity and displacement of the journal. The deformation of the foil is evaluated through numerical iteration in feedback mode with the fluid film pressure generated by the journal motion until convergence at every time step is achieved. For the full transient simulation, a parallel real-time integration scheme is used to evaluate simultaneously the new journal position and the new deformed shape of the foil at each time step. The pressure of the fluid lubricant is iterated jointly with the corresponding journal position and the deformed foil geometry until convergence is achieved. A variable time-stepping Newmark-Beta integration procedure is used to evaluate the transient dynamics at each time step of the bearing.

Braun, M. Jack↗
NASA NTRS1996

ARTSN: An Automated Real-Time Spacecraft Navigation System

As part of the Deep Space Network (DSN) advanced technology program an effort is underway to design a filter to automate the deep space navigation process.The automated real-time spacecraft navigation (ARTSN) filter task is based on a prototype consisting of a FORTRAN77 package operating on an HP-9000/700 workstation running HP-UX 9.05. This will be converted to C, and maintained as the operational version. The processing tasks required are: (1) read a measurement, (2) integrate the spacecraft state to the current measurement time, (3) compute the observable based on the integrated state, and (4) incorporate the measurement information into the state using an extended Kalman filter. This filter processes radiometric data collected by the DSN. The dynamic (force) models currently include point mass gravitational terms for all planets, the Sun and Moon, solar radiation pressure, finite maneuvers, and attitude maintenance activity modeled quadratically. In addition, observable errors due to troposphere are included. Further data types, force and observable models will be ncluded to enhance the accuracy of the models and the capability of the package. The heart of the ARSTSN is a currently available continuous-discrete extended Kalman filter. Simulated data used to test the implementation at various stages of development and the results from processing actual mission data are presented.

Burkhart, P. Daniel↗
NASA NTRS1996

Conversion-Integration of MSFC Nonlinear Signal Diagnostic Analysis Algorithms for Realtime Execution of MSFC's MPP Prototype System

NASA's advanced propulsion system Small Scale Magnetic Disturbances/Advanced Technology Development (SSME/ATD) has been undergoing extensive flight certification and developmental testing, which involves large numbers of health monitoring measurements. To enhance engine safety and reliability, detailed analysis and evaluation of the measurement signals are mandatory to assess its dynamic characteristics and operational condition. Efficient and reliable signal detection techniques will reduce the risk of catastrophic system failures and expedite the evaluation of both flight and ground test data, and thereby reduce launch turn-around time. During the development of SSME, ASRI participated in the research and development of several advanced non- linear signal diagnostic methods for health monitoring and failure prediction in turbomachinery components. However, due to the intensive computational requirement associated with such advanced analysis tasks, current SSME dynamic data analysis and diagnostic evaluation is performed off-line following flight or ground test with a typical diagnostic turnaround time of one to two days. The objective of MSFC's MPP Prototype System is to eliminate such 'diagnostic lag time' by achieving signal processing and analysis in real-time. Such an on-line diagnostic system can provide sufficient lead time to initiate corrective action and also to enable efficient scheduling of inspection, maintenance and repair activities. The major objective of this project was to convert and implement a number of advanced nonlinear diagnostic DSP algorithms in a format consistent with that required for integration into the Vanderbilt Multigraph Architecture (MGA) Model Based Programming environment. This effort will allow the real-time execution of these algorithms using the MSFC MPP Prototype System. ASRI has completed the software conversion and integration of a sequence of nonlinear signal analysis techniques specified in the SOW for real-time execution on MSFC's MPP Prototype. This report documents and summarizes the results of the contract tasks; provides the complete computer source code; including all FORTRAN/C Utilities; and all other utilities/supporting software libraries that are required for operation.

Jong, Jen-Yi↗
NASA NTRS1997

Condensate Accretion in Shock Tube's Expansion Fan

It has been shown that turbulence and temperature influence the droplet sizes in expansion fan induced condensation by studying the Rayleigh scattering from one port in our shock tube's test section. We have modified our set-up so as to allow, using two ports, the real time measurement of the influence of turbulence and temperature on the rate at which these droplets grow. To do this, we looked at the Rayleigh scattering from two different ports for ten Reynolds numbers at five different temperatures. We modeled the time of flight of droplets, using the equations of one-dimensional gas dynamics and the measured shock wave speed in shock tube's driven section.

Mezonlin, Ephrem-Denis↗
NASA NTRS1998

Time-Resolved O3 Chemical Chain Reaction Kinetics Via High-Resolution IR Laser Absorption Methods

Excimer laser photolysis in combination with time-resolved IR laser absorption detection of OH radicals has been used to study O3/OH(v = 0)/HO2 chain reaction kinetics at 298 K, (i.e.,(k(sub 1) is OH + 03 yields H02 + 02 and (k(sub 2) is H02 + 03 yields OH + 202). From time-resolved detection of OH radicals with high-resolution near IR laser absorption methods, the chain induction kinetics have been measured at up to an order of magnitude higher ozone concentrations ([03] less than or equal to 10(exp 17) molecules/cu cm) than accessible in previous studies. This greater dynamic range permits the full evolution of the chain induction, propagation, and termination process to be temporally isolated and measured in real time. An exact solution for time-dependent OH evolution under pseudo- first-order chain reaction conditions is presented, which correctly predicts new kinetic signatures not included in previous OH + 03 kinetic analyses. Specifically, the solutions predict an initial exponential loss (chain "induction") of the OH radical to a steady-state level ([OH](sub ss)), with this fast initial decay determined by the sum of both chain rate constants, k(sub ind) = k(sub 1) + k(sub 2). By monitoring the chain induction feature, this sum of the rate constants is determined to be k(sub ind) = 8.4(8) x 10(exp -14) cu cm/molecule/s for room temperature reagents. This is significantly higher than the values currently recommended for use in atmospheric models, but in excellent agreement with previous results from Ravishankara et al.

Kulcke, Axel↗
NASA NTRS2000

Towards a Comprehensive Dynamic-chemistry Assimilation for Eos-Chem: Plans and Status in NASA's Data Assimilation Office

In order to support the EOS-Chem project, a comprehensive assimilation package for the coupled chemical-dynamical system is being developed by the Data Assimilation Office at NASA GSFC. This involves development of a coupled chemistry/meteorology model and of data assimilation techniques for trace species and meteorology. The model is being developed using the flux-form semi-Lagrangian dynamical core of Lin and Rood, the physical parameterizations from the NCAR Community Climate Model, and atmospheric chemistry modules from the Atmospheric Chemistry and Dynamics branch at NASA GSFC. To date the following results have been obtained: (i) multi-annual simulations with the dynamics-radiation model show the credibility of the package for atmospheric simulations; (ii) initial simulations including a limited number of middle atmospheric trace gases reveal the realistic nature of transport mechanisms, although there is still a need for some improvements. Samples of these results will be shown. A meteorological assimilation system is currently being constructed using the model; this will form the basis for the proposed meteorological/chemical assimilation package. The latter part of the presentation will focus on areas targeted for development in the near and far terms, with the objective of Providing a comprehensive assimilation package for the EOS-Chem science experiment. The first stage will target ozone assimilation. The plans also encompass a reanalysis (ReSTS) for the 1991-1995 period, which includes the Mt. Pinatubo eruption and the time when a large number of UARS observations were available. One of the most challenging aspects of future developments will be to couple theoretical advances in tracer assimilation with the practical considerations of a real environment and eventually a near-real-time assimilation system.

Pawson, Steven↗
NASA NTRS2000

Nozomi Cis-Lunar Phase Orbit Determination

Japan's Institute of Space and Astronautical Science (ISAS) launched Nozomi, its first mission to the planet Mars using the newly developed M-V launch vehicle on July 3, 1998. Scientific objectives of the mission are to study the structure and dynamics of the Martian upper atmosphere and its interaction with the solar wind. Nozomi is a cooperative mission between ISAS and the National Aeronautics and Space Administration (NASA). The NASA contribution includes navigation and tracking services provided by the Jet Propulsion Laboratory (JPL). The spacecraft also serves as an engineering demonstration of basic technology for planetary exploration. One of the new technologies was a unique trajectory, developed by ISAS, which used solar gravitational perturbations at the weak stability boundary as an aid to achieve an Earth-Mars transfer orbit. This trajectory saves approximately 120 m/s of Delta V compared to direct hyperbolic insertion and is considered an enabling technology for the mission. Nozomi was the first spacecraft to employ this trajectory and provided on-orbit validation of the technique. The trajectory was achieved by initially placing the spacecraft in a highly elliptical cis-lunar phasing orbit. Six maneuvers were performed during this period to correct injection errors and target an outbound lunar swingby in September 1998. The gravity assist from the lunar swingby raised apogee to the vicinity of the weak stability boundary. After three more targeting maneuvers, Nozomi performed an inbound lunar swingby followed immediately by a powered Earth swingby in late December 1998. A 420 m/s Trans Mars Insertion (TMI) burn at the final Earth periapsis was intended to place the spacecraft on a heliocentric trajectory leading to Mars orbit insertion in October 1999. Orbit determination for Nozomi is performed in parallel by both ISAS and the Multi-Mission Navigation (MMNAV) group at JPL. This was an advantage for the mission because each group would generate solutions based on data collected from their respective tracking networks. Spacecraft events, such as sequence uplinks and maneuvers, were generally scheduled during passes at the Usuda tracking station in Japan. As a result, maneuver design and reconstruction was derived from MMNAV solutions based on JPL tracking data obtained immediately prior to or following maneuvers. Data was also exchanged between ISAS and MMNAV so orbit determination could be performed on joint data sets in support of critical targeting late in the cis-lunar phase. In this paper, information regarding the MMNAV orbit determination effort for the first six months of the mission is presented. The spacecraft trajectory is characterized first, followed by a discussion of the orbit determination estimation procedure and models. Results from selected orbit solutions are presented and compared against reconstructed trajectories. One area of emphasis in this paper is orbit determination in the vicinity of the weak stability boundary. Precise navigation was necessary to target the second lunar swingby and the powered Earth swingby. Delivery accuracy of 150 m was required for these critical encounters, but a number of factors contributed to the general degradation of orbit determination accuracy. This included the fact that the spacecraft was at apogee, at a range of 1.7 million km and moving at less than I km/sec perpendicular to the line of sight. Nozomi was also close to zero degrees declination where there are known limitations on orbit determination performance. Finally, S-band tracking data was acquired through the Nozomi backup low gain antenna. This antenna is offset from the axis of this spin stabilized spacecraft and superimposed large signatures in the Doppler and range data. These difficulties were overcome by combining long data arcs, spanning several maneuvers, with a high fidelity solar pressure model. The model included a physically accurate representation of the spacecraft structure and a high time resolution orientation model. Observation modeling included the removal of the spin induced Doppler bias, spin signature and per pass correction of range calibration errors applied for data leading up to critical events. As a result, all orbit determination goals were met. A second area of emphasis in this paper is the JPL tracking and orbit determination effort in support of the TMI maneuver. TMI occurred out of contact with ground stations and the JPL Goldstone tracking complex had the first pass following the bum. As a result, MMNAV had the responsibility to make a rapid assessment of the maneuver performance. MMNAV made the determination that a 100 m/s under bum had occurred and promptly informed ISAS via voice lines. ISAS immediately began preparations for a correction maneuver (TMIc), which had to be performed during the next Usuda pass. The near real time assessment by MMNAV provided accurate antenna frequency and pointing updates for the spacecraft acquisition at Usuda and the close coordination between the two agencies enabled the design and successful execution of the TMc maneuver. Propellant consumption during the correction burn dictated that the mission be redesigned. ISAS developed a new plan which adds 3 full solar orbits, two Earth swingbys and one lunar swingby with arrival at Mars in January 2004. The final Mars orbit will still enable the mission to achieve all of its science objectives.

Ryne, Mark↗
NASA NTRS2000

On-Board Real-Time State and Fault Identification for Rovers

For extended autonomous operation, rovers must identify potential faults to determine whether its execution needs to be halted or not. At the same time, rovers present particular challenges for state estimation techniques: they are subject to environmental influences that affect senior readings during normal and anomalous operation, and the sensors fluctuate rapidly both because of noise and because of the dynamics of the rover's interaction with its environment. This paper presents MAKSI, an on-board method for state estimation and fault diagnosis that is particularly appropriate for rovers. The method is based on a combination of continuous state estimation, wing Kalman filters, and discrete state estimation, wing a Markov-model representation.

Washington, Richard↗
NASA NTRS2002

Mars Smart Lander Simulations for Entry, Descent, and Landing

Two primary simulations have been developed and are being updated for the Mars Smart Lander Entry, Descent, and Landing (EDL). The high fidelity engineering end-to-end EDL simulation that is based on NASA Langley's Program to Optimize Simulated Trajectories (POST) and the end-to-end real-time, hardware-in-the-loop simulation testbed, which is based on NASA JPL's (Jet Propulsion Laboratory) Dynamics Simulator for Entry, Descent and Surface landing (DSENDS). This paper presents the status of these Mars Smart Lander EDL end-to-end simulations at this time. Various models, capabilities, as well as validation and verification for these simulations are discussed.

Striepe, S. A.↗
NASA NTRS2002

Adaptive Identification and Control of Flow-Induced Cavity Oscillations

Progress towards an adaptive self-tuning regulator (STR) for the cavity tone problem is discussed in this paper. Adaptive system identification algorithms were applied to an experimental cavity-flow tested as a prerequisite to control. In addition, a simple digital controller and a piezoelectric bimorph actuator were used to demonstrate multiple tone suppression. The control tests at Mach numbers of 0.275, 0.40, and 0.60 indicated approx. = 7dB tone reductions at multiple frequencies. Several different adaptive system identification algorithms were applied at a single freestream Mach number of 0.275. Adaptive finite-impulse response (FIR) filters of orders up to N = 100 were found to be unsuitable for modeling the cavity flow dynamics. Adaptive infinite-impulse response (IIR) filters of comparable order better captured the system dynamics. Two recursive algorithms, the least-mean square (LMS) and the recursive-least square (RLS), were utilized to update the adaptive filter coefficients. Given the sample-time requirements imposed by the cavity flow dynamics, the computational simplicity of the least mean squares (LMS) algorithm is advantageous for real-time control.

Kegerise, M. A.↗
NASA NTRS2003

Multiple Input Design for Real-Time Parameter Estimation in the Frequency Domain

A method for designing multiple inputs for real-time dynamic system identification in the frequency domain was developed and demonstrated. The designed inputs are mutually orthogonal in both the time and frequency domains, with reduced peak factors to provide good information content for relatively small amplitude excursions. The inputs are designed for selected frequency ranges, and therefore do not require a priori models. The experiment design approach was applied to identify linear dynamic models for the F-15 ACTIVE aircraft, which has multiple control effectors.

Morelli, Eugene↗
NASA NTRS2002

Dynamical Studies of the Middle Atmosphere Using High Resolution Doppler Imager Observations

This report summarizes the activities of NASA grant NAG5-11068, "Dynamicai Studies of the Middle Atmosphere Using High Resolution Doppler Imager Observations." The High Resolution Doppler Imager (HRDI) on the Upper Atmosphere Research Satellite (UARS) has been providing direct measurements of the Earth's horizontal wind field in the stratosphere, mesosphere and lower thermosphere. Mesospheric temperatures, ozone, and O((sup 1) D) densities, and stratospheric aerosol extinctions coefficients, are also retrieved. The goal of HRDI is to measure the vector winds in the stratosphere (10-40 km), mesosphere, and lower thermosphere (approximately 50-120 km) during the day, and the lower thermosphere at night (approximately 95 km) to an accuracy of 5 m/s. The horizontal wind vector is measured by observing the Doppler shift of rotational lines of molecular oxygen along two lines of sight. In addition to winds, temperatures and volume emission rates are determined in the mesosphere and lower thermosphere, from which ozone and O((sup 1) D) concentrations can be derived, and aerosol scattering coefficients are determined in the stratosphere. UARS was launched on September 12, 1991, into a 585-km circular orbit inclined 57 degrees to the equator HRDI was activated September 28, 1991 and following a period of checkout and adjustment of the instrument parameters, scientific observations began November 2, 199 1. HRDI operated nearly continuously from launch until April 1995. At that time the UARS solar array drive failed, forcing the instruments to time-share the available power. From July 1995 to July 1996 HRDI operated approximately 50% of the time. At that point, one of the three spacecraft batteries failed and from then until September 1998 the duty cycle was less than 20% per month, At that time it was determined that HRDI could operate during each daytime pass, which increased the daytime duty cycle to close to l00%, while nighttime operations were limited to about a week per month. In the fall of 1999, the second tape recorder failed requiring a real time contact with a TRDSS satellite to retrieve that data. This resulted in about 60% data collection efficiency. Finally, in the summer of 2000, the second star sensor failed requiring the spacecraft attitude to be controlled by a three axis magnetometer and sun sensor. This resulted in a loss of attitude knowledge but operations continue with the anticipation of correcting the attitude. A new method for determining the tide and mean structure from satellite data in conjunction with a new tidal model has been devised. For brevity, it shall be referred to as the TMAT or Tide-Mean Assimilation Technique. Most previous methods of tidal analysis are based on various ways of slicing the data set.

Skinner, Wilbert↗
DOE OSTI2025

Nature-GL: A Revolutionary Learning Paradigm Unleashing Nature’s Power in Real-World Spatial-Temporal Graph Learning

Spatial-Temporal Graph Learning (ST-GL) is a prominent research area due to its unique capability to effectively learn real-world graphs. Applications of ST-GL pose stringent and various demands on not only real-time inference with low energy cost and high ac- curacy but also fast training. Unfortunately, as Moore’s Law approaches its limits and ST-GL model complexity drastically grows, the gap between digital hardware’s computational power and ST- GL application demands is widening. In response, this paper introduces Nature-GL, a nature-powered graph learning paradigm that exploits the principle of entropy increase to advance graph learning. In particular, Nature-GL transforms both the training and inference of real-valued ST-GL into electron-speed natural anneal- ing processes of a parameterized dynamical system that represents the target graphs. Experimental results across four real-world ap- plications with six datasets demonstrate that Nature-GL achieves orders-of-magnitude speedups in both training and inference, delivering higher accuracy compared to Graph Neural Networks.

Liu, Chuan [University of Rochester]↗
DOE OSTI2025

Predicting Initial Trans-Membrane Pressure for Optimized Operations in UF Unit Using Random Forest

With the growing scarcity of freshwater, innovative process design mechanisms like Reverse Osmosis (RO) are increasingly gaining attention among water treatment utilities to address the rising demand. Ensuring reliable water production necessitates efficient resource utilization, minimizing downtime in (ultra-filtration) UF systems. Recent advancements in machine learning (ML) have enabled the development of accurate data-driven models for Model Predictive Control (MPC), often requiring minimal prior knowledge of underlying physical processes. In this study, we present predictive regression models based on Random Forest (RF) and Auto-Regressive (AR) approaches to forecast the initial Trans-Membrane Pressure (TMP) for each filtration cycle in data generated by Direct Potable Reuse (DPR) systems. The proposed RF-based model demonstrates superior performance compared to baseline methods, including historical mean, Last Observation Carried Forward (LOCF), and naïve AR models, across various forecasting horizons in terms of root mean square error (RMSE) metric. To evaluate how different classes of process variables contribute to TMP dynamics over time, we examine the feature importance of independent covariates across multiple forecast horizons. This analysis provides insight into the temporal relevance of operational and sensor-derived features, guiding control and monitoring strategies. Additionally, the impact of hyperparameter tuning on TMP prediction performance is studied for both direct and recursive RF modelling approaches across increasing forecast horizons. Accurate prediction of initial TMP is critical for optimizing RO operations, as it enables the development of robust modelling frameworks by accurately estimating membrane fouling trends, thereby enhancing process efficiency and long-term reliability. The demonstrated efficacy of the RF-based approach highlights its potential as a tool for real-time decision-making in water treatment systems, paving the way for advanced process optimization and sustainable water resource management.

Mukherjee, Subrata [ORNL] (ORCID:0000000309930338)↗