Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model optimization”

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 919 records · Page 51
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Once-Through Steam Generator Model Analysis Using Python and Advanced Optimization Tools (Summer Internship Report)

This study focuses on the parametric analysis of design parameters for a once-through steam generator (OTSG) model, using python and advanced optimization tools to facilitate applications such as the flowing autoclave steam generator (FASG) test cases. Building on previous research involving another OTSG with a different design, this project aims to enhance our understanding of how steam generators (SGs) behave and how their outputs are influenced by changes in design. The reason for this design change is to allow for more precise modeling and optimization of SG performance, to provide a comparative analysis between the two designs, and to set up the model for integration with the FASG test case. The OTSG python-model is a mathematical representation (including fluid flow and heat transfer equations/models/correlations) of a steam-generating unit in a pressurized water reactor-type small modular reactor system. Design studies involve changing the model’s input design parameters to observe the resulting effects on the output of the system. By using advanced optimization tools, such as the Risk Analysis Virtual Environment (RAVEN) developed at Idaho National Laboratory, detailed design parametric studies and model optimization were performed. Six input parameters—pressure, temperature, and mass flow rate for the inlet of the primary-side (hot fluid) and secondary-side (cold fluid), respectively, of the SG—were randomly perturbed via RAVEN’s Monte Carlo Sampler module, using uniform distributions (i.e., ±1%, ±5% and ±10% relative changes) for 600 samples. The analysis provides valuable insights into SG optimization and can be used for sensor placement optimization to effectively monitor and obtain experimental data in other tests.

20 FOSSIL-FUELED POWER PLANTS↗
NASA NTRS2001

A Simplified Model of ARIS for Optimal Controller Design

Many space-science experiments require active vibration isolation. Boeing's Active Rack Isolation System (ARIS) isolates experiments at the rack (vs. experiment or sub-experiment) level, with multi e experiments per rack. An ARIS-isolated rack typically employs eight actuators and thirteen umbilicals; the umbilicals provide services such as power, data transmission, and cooling. Hampton, et al., used "Kane's method" to develop an analytical, nonlinear, rigid-body model of ARIS that includes full actuator dynamics (inertias). This model, less the umbilicals, was first implemented for simulation by Beech and Hampton; they developed and tested their model using two commercial-off-the-shelf (COTS) software packages. Rupert, et al., added umbilical-transmitted disturbances to this nonlinear model. Because the nonlinear model, even for the untethered system, is both exceedingly complex and "encapsulated" inside these COTS tools, it is largely inaccessible to ARIS controller designers. This paper shows that ISPR rattle-space constraints and small ARIS actuator masses permit considerable model simplification, without significant loss of fidelity. First, for various loading conditions, comparisons are made between the dynamic responses of the nonlinear model (untethered) and a truth model. Then comparisons are made among nonlinear, linearized, and linearized reduced-mass models. It is concluded that these three models all capture the significant system rigid-body dynamics, with the third being preferred due to its relative simplicity.

Beech, Geoffrey S.↗
DOE OSTI2023

Applied Machine-Learning Models to Identify Spectral Sub-Types of M Dwarfs from Photometric Surveys

M dwarfs are the most abundant stars in the Solar Neighborhood and they are prime targets for searching for rocky planets in habitable zones. Consequently, a detailed characterization of these stars is in demand. The spectral sub-type is one of the parameters that is used for the characterization and it is traditionally derived from the observed spectra. However, obtaining the spectra of M dwarfs is expensive in terms of observation time and resources due to their intrinsic faintness. We study the performance of four machine-learning (ML) models—K-Nearest Neighbor (KNN), Random Forest (RF), Probabilistic Random Forest (PRF), and Multilayer Perceptron (MLP)—in identifying the spectral sub-types of M dwarfs at a grand scale by deploying broadband photometry in the optical and near-infrared. We trained the ML models by using the spectroscopically identified M dwarfs from the Sloan Digital Sky Survey (SDSS) Data Release (DR) 7, together with their photometric colors that were derived from the SDSS, Two-Micron All-Sky Survey, and Wide-field Infrared Survey Explorer. We found that the RF, PRF, and MLP give a comparable prediction accuracy, 74%, while the KNN provides slightly lower accuracy, 71%. We also found that these models can predict the spectral sub-type of M dwarfs with ~99% accuracy within ±1 sub-type. The five most useful features for the prediction are r - z, r - i, r - J, r - H , and g - z, and hence lacking data in all SDSS bands substantially reduces the prediction accuracy. However, we can achieve an accuracy of over 70% when the r and i magnitudes are available. Since the stars in this study are nearby (d ≲ 1300 pc for 95% of the stars), the dust extinction can reduce the prediction accuracy by only 3%. Finally, we used our optimized RF models to predict the spectral sub-types of M dwarfs from the Catalog of Cool Dwarf Targets for the Transiting Exoplanet Survey Satellite, and we provide the optimized RF models for public use.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Machine Learning Benchmarks for the Classification of Equivalent Circuit Models from Electrochemical Impedance Spectra

Analysis of Electrochemical Impedance Spectroscopy (EIS) data for electrochemical systems often consists of defining an Equivalent Circuit Model (ECM) using expert knowledge and then optimizing the model parameters to deconvolute various resistance, capacitive, inductive, or diffusion responses. For small data sets, this procedure can be conducted manually; however, it is not feasible to manually define a proper ECM for extensive data sets with a wide range of EIS responses. Automatic identification of an ECM would substantially accelerate the analysis of large sets of EIS data. We showcase machine learning methods to classify the ECMs of 9,300 impedance spectra provided by QuantumScape for the BatteryDEV hackathon. The best-performing approach is a gradient-boosted tree model utilizing a library to automatically generate features, followed by a random forest model using the raw spectral data. A convolutional neural network using boolean images of Nyquist representations is presented as an alternative, although it achieves a lower accuracy. We publish the data and open source the associated code. The approaches described in this article can serve as benchmarks for further studies. A key remaining challenge is the identifiability of the labels, underlined by the model performances and the comparison of misclassified spectra.

25 ENERGY STORAGE↗
DOE OSTI2026

Enhancing segmentation fairness through curriculum learning and progressive loss: a centralized and federated perspective on radiograph analysis

Bias in medical image segmentation can lead to unequal performance across demographic subgroups, raising concerns about fairness and reliability in clinical AI systems. While deep learning models have achieved high segmentation accuracy, ensuring equitable performance across race and gender remains a significant challenge, particularly in privacy-sensitive healthcare environments. This study investigates fairness-aware medical image segmentation for hip and knee radiographs using deep learning models evaluated in both centralized and Federated Learning (FL) settings. We introduce Curriculum Learning (CL) strategies and Progressive Loss (PL) functions to regulate sample difficulty during training. In addition, we propose two novel fairness-oriented federated learning algorithms, Federated Intersection over Union (FedIoU) and Federated Intersection over Union with Outlier Analysis (FedIoUoutlier). Experiments are conducted using multiple segmentation backbones and simulated multi-site data partitions derived from the Osteoarthritis Initiative dataset. Model performance is evaluated using Intersection over Union (IoU), IoU standard deviation, Skewed Error Ratio (SER), and Min-Max Disparity across race and gender subgroups. Statistical significance was verified using paired t-tests to compare per-sample IoU performance against baseline configurations. Across both hip and knee segmentation tasks, curriculum learning and progressive loss strategies consistently improved segmentation accuracy and reduced demographic performance disparities in centralized training. In federated settings, fairness-aware aggregation further enhanced performance. Notably, FedIoUoutlier combined with balanced curriculum learning and tiered progressive loss achieved the highest mean IoU while yielding the lowest SER and Min-Max Disparity, indicating improved fairness without sacrificing accuracy. In several configurations, federated models matched or exceeded the performance of optimized centralized models, with statistically significant improvements in per-sample IoU over baseline configurations. The results demonstrate that structured training strategies and fairness-aware federated aggregation can jointly improve accuracy, stability, and demographic fairness in medical image segmentation. By integrating curriculum learning, progressive loss, and novel FL algorithms, this work provides a practical pathway toward equitable and privacy-preserving AI systems for medical imaging.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1980

Closed loop models for analyzing engineering requirements for simulators

A closed loop analytic model, incorporating a model for the human pilot, (namely, the optimal control model) that would allow certain simulation design tradeoffs to be evaluated quantitatively was developed. This model was applied to a realistic flight control problem. The resulting model is used to analyze both overall simulation effects and the effects of individual elements. The results show that, as compared to an ideal continuous simulation, the discrete simulation can result in significant performance and/or workload penalties.

Baron, S.↗
NASA NTRS1980

Handling qualities of large flexible control-configured aircraft

The effects on handling qualities of low frequency symmetric elastic mode interaction with the rigid body dynamics of a large flexible aircraft was analyzed by use of a mathematical pilot modeling computer simulation. An extension of the optimal control model for a human pilot was made so that the mode interaction effects on the pilot's control task could be assessed. Pilot ratings were determined for a longitudinal tracking task with parametric variations in the undamped natural frequencies of the two lowest frequency symmetric elastic modes made to induce varying amounts of mode interaction. Relating numerical performance index values associated with the frequency variations used in several dynamic cases, to a numerical Cooper-Harper pilot rating has proved successful in discriminating when the mathematical pilot can or cannot separate rigid from elastic response in the tracking task.

Swaim, R. L.↗
NASA NTRS1982

Modeling human target acquisition in ground-to-air weapon systems

The problems associated with formulating and validating mathematical models for describing and predicting human target acquisition response are considered. In particular, the extension of the human observer model to include the acquisition phase as well as the tracking segment is presented. Relationship of the Observer model structure to the more complex Standard Optimal Control model formulation and to the simpler Transfer Function/Noise representation is discussed. Problems pertinent to structural identifiability and the form of the parameterization are elucidated. A systematic approach toward the identification of the observer acquisition model parameters from ensemble tracking error data is presented.

Phatak, A. V.↗
NASA NTRS2014

Assessing the Risk of Crew Injury Due to Dynamic Loads During Spaceflight

Spaceflight requires tremendous amounts of energy to achieve Earth orbit and to attain escape velocity for interplanetary missions. Although the majority of the energy is managed in such a way as to limit the accelerations on the crew, several mission phases may result in crew exposure to dynamic loads. In the automotive industry, risk of serious injury can be tolerated because the probability of a crash is remote each time a person enters a vehicle, resulting in a low total risk of injury. For spaceflight, the level of acceptable injury risk must be lower to achieve a low total risk of injury because the dynamic loads are expected on each flight. To mitigate the risk of injury due to dynamic loads, the NASA Human Research Program has developed a research plan to inform the knowledge gaps and develop relevant tools for assessing injury risk. The risk of injury due to dynamic loads can be further subdivided into extrinsic and intrinsic risk factors. Extrinsic risk factors include the vehicle dynamic profile, seat and restraint design, and spacesuit design. Human tolerance to loads varies considerably depending on the direction, amplitude, and rise-time of acceleration therefore the orientation of the body to the dynamic vector is critical to determining crew risk of injury. Although a particular vehicle dynamic profile may be safe for a particular design, the seat, restraint, and suit designs can affect the risk of injury due to localized effects. In addition, characteristics intrinsic to the crewmember may also contribute to the risk of injury, such as crewmember sex, age, anthropometry, and deconditioning due to spaceflight, and each astronaut may have a different risk profile because of these factors. The purpose of the research plan is to address any knowledge gaps in the risk factors to mitigate injury risk. Methods for assessing injury risk have been well documented in other analogous industries and include human volunteer testing, human exposure to dynamic environments, post-mortem human subject (PMHS) testing, animal testing, anthropomorphic test devices (ATD), dynamic models of the human, numerical models of ATDs, and numerical models of the human. Each has inherent strengths and limitations. For example, human volunteer testing is advantageous because a population can be selected that is similar to the astronaut corps; however, because of the inherent ethical limitations, only sub-injurious conditions can be tested. PMHSs can be tested in a variety of conditions including injurious levels, but the responses are not completely analogous to living human subjects. In addition, it is exceedingly difficult to select a PMHS population that is similar to the astronaut corps. ATDs are currently widely used in the automotive industry and military because they are highly repeatable and durable. Unfortunately, because they are mechanical models of the human body, the biofidelity of the responses are limited to dynamic conditions used to validate the ATD. Numerical models of the ATD, in addition to the strengths and limitations for ATDs, are easy to use for a variety of designs before a design is fabricated, but also have additional limitations for ATDs, are easy to use for a variety of designs before a design is fabricated, but also have additional uncertainty. Dynamic models are simple and easy to use, but do not account for localized effects of the seat and suit. Finally, numerical models of the human have the potential to have the most advantages; however, the current models are not validated for the conditions expected during spaceflight. To properly assess spaceflight conditions with numerical human models, human data would be needed to optimize the model responses for those conditions. Using the appropriate assessment method with the knowledge gained for each risk factor, an appropriate approach for mitigating the risk of injury due to dynamic loads can be developed ensuring crew safety in future NASA vehicles.

Somers, J. T.↗
NASA NTRSDate not supplied

Modeling and Analysis of Stirling Power Convertors

Modeling and Analysis of Stirling Power Convertors Luis A. Rodriguez1 Steven M. Geng, Terry V. Reid, Scott D. Wilson NASA Glenn Research Center, Cleveland, OH, 44135, USA NASA Glenn’s Thermal Energy Conversion Branch is supporting the development of the next generation free-piston Stirling power convertors. American Superconductor (AMSC) and Sunpower Inc. are the two firms under contract to develop the Flexure Isotope Stirling Convertor (FISC) and the Sunpower Robust Stirling Convertor (SRSC), respectively. To comprehend and forecast convertor performance, Sage, ANSYS® Maxwell, and ANSYS® Fluent were used to model the Stirling thermodynamic cycle, alternator electromagnetics, and piston and displacer dynamics. I. Introduction Stirling convertors are being developed by NASA as a potential steady source of electrical power for NASA’s future scientific space missions. Currently, NASA Glenn Research Center has two corporations under contract, American Superconductor (AMSC) and Sunpower Inc., for the development of the next generation of free-piston Stirling convertors for dynamic radioisotope power systems. AMSC is developing the Flexure Isotope Stirling Convertor (FISC), which uses flexures to prevent side motion and rubbing of the piston. Similarly, Sunpower Inc, is developing the Sunpower Robust Stirling Convertor (SRSC). The SRSC uses gas bearings to prevent radial contact of the moving piston. As convertor development continues, it is increasingly important to understand and predict the interactions of components in the system, how they respond to one another, and how they perform as a response to changes in operating conditions. A suitable and enlightening way to demonstrate and foresee these interactions is with the use of accurate modeling software. Sage, ANSYS® Maxwell, and ANSYS® Fluent are the current modeling tools used by NASA to analytically determine convertor performance. Sage is a one-dimensional object-oriented commercial software package used for modeling and optimizing Stirling convertors for Dynamic Radioisotope Power Systems (DRPS) and it is one of the most accurate Stirling convertor codes in use by NASA. This code is the successor to GLIMPS (Globally-Implicit Stirling Cycle Simulation) and GLOP (GLIMPS Optimization) software created by Gedeon Associates [1]. Model input parameters are typically material/gas type, component physical dimensions, temperatures, frequency, charge pressure, and number of time/space nodes. Sage is used to model both the FISC’s and SRSC’s Stirling cycle thermodynamics and piston/displacer dynamics. Performance maps were created and analyzed for both power systems to better understand the relationship between the following conditions: cold-end temperature, hot-end temperature, piston/displacer amplitudes, pressure drop, and thermal input power. The synergy between these conditions will help determine parameter sensitivity. ANSYS® Maxwell was used to create a three-dimensional (3-D) axisymmetric model for both FISC and SRSC alternators. The significant physical components included in each model are the magnets, magnet carrier, outer/inner laminations, and the coil. Inputs to the model are piston amplitude, piston frequency, alternator load, coil resistance, tuning capacitance, and specific material properties. The alternator models calculate terminal voltage, current, piston/current phase, voltage/current phase, coil inductance, terminal power and efficiency. The RI2 losses, core (hysteresis and eddy) losses, and magnet/can eddy losses are also a part of the final results. ANSYS® Fluent is used to build 3-D computational fluid dynamic (CFD) models to examine the Stirling cycle thermodynamics for both the FISC and SRSC systems. Three-dimensional Computer Aided Design (CAD) models were used to create the physical components of each convertor. Steady-state simulations were conducted for hardware testing, prediction of environmental losses during testing, and generation of radiation look-up tables. The model inputs to the aforementioned analysis are the material properties and boundary thermal conditions. The steady-state model calculates temperature and heat flow distributions. Transient 3-D calculations were also part of the CFD analysis. In this study a physically reduced version of the FISC is used to obtain a prediction of available engine power. For the gas bearing SRSC, the transient effort is used to obtain a prediction of bearing pad performance and its sensitivity to micro-channel geometric variation. The model inputs to the transient simulations are the piston amplitude, displacer amplitude, frequency, displacer/piston phase angle, dynamic deforming CFD grid, temperature boundary conditions, and user defined files describing motion profile of piston/displacer. The results of the model are temperature distributions, heat distributions, and PV power produced at pre-determined conditions.

Luis A Rodriguez↗
NASA NTRS1995

Performance seeking control (PSC) for the F-15 highly integrated digital electronic control (HIDEC) aircraft

The performance seeking control algorithm optimizes total propulsion system performance. This adaptive, model-based optimization algorithm has been successfully flight demonstrated on two engines with differing levels of degradation. Models of the engine, nozzle, and inlet produce reliable, accurate estimates of engine performance. But, because of an observability problem, component levels of degradation cannot be accurately determined. Depending on engine-specific operating characteristics PSC achieves various levels performance improvement. For example, engines with more deterioration typically operate at higher turbine temperatures than less deteriorated engines. Thus when the PSC maximum thrust mode is applied, for example, there will be less temperature margin available to be traded for increasing thrust.

Orme, John S.↗
NASA NTRS1990

A technique for optimal temperature estimation for modeling sunrise/sunset thermal snap disturbance torque

A predictive temperature estimation technique which can be used to drive a model of the Sunrise/Sunset thermal 'snap' disturbance torque experienced by low Earth orbiting spacecraft is described. The twice per orbit impulsive disturbance torque is attributed to vehicle passage in and out of the Earth's shadow cone (umbra), during which large flexible appendages undergo rapidly changing thermal conditions. Flexible members, in particular solar arrays, experience rapid cooling during umbra entrance (Sunset) and rapid heating during exit (Sunrise). The thermal 'snap' phenomena has been observed during normal on-orbit operations of both the LANDSAT-4 satellite and the Communications Technology Satellite (CTS). Thermal 'snap' has also been predicted to be a dominant source of error for the TOPEX satellite. The fundamental equations used to model the Sunrise/Sunset thermal 'snap' disturbance torque for a typical solar array like structure will be described. For this derivation the array is assumed to be a thin, cantilevered beam. The time varying thermal gradient is shown to be the driving force behind predicting the thermal 'snap' disturbance torque and therefore motivates the need for accurate estimates of temperature. The development of a technique to optimally estimate appendage surface temperature is highlighted. The objective analysis method used is structured on the Gauss-Markov Theorem and provides an optimal temperature estimate at a prescribed location given data from a distributed thermal sensor network. The optimally estimated surface temperatures could then be used to compute the thermal gradient across the body. The estimation technique is demonstrated using a typical satellite solar array.

Zimbelman, D. F.↗
DOE OSTI2022

An improved representation of the relationship between photosynthesis and stomatal conductance leads to more stable estimation of conductance parameters and improves the goodness-of-fit across diverse data sets

Stomata play a central role in surface-atmosphere exchange by controlling the flux of water and CO 2 between the leaf and the atmosphere. Representation of stomatal conductance (g sw ) is therefore an essential component of models that seek to simulate water and CO 2 exchange in plants and ecosystems. For given environmental conditions at the leaf surface (CO 2 concentration and vapor pressure deficit or relative humidity), models typically assume a linear relationship between g sw and photosynthetic CO 2 assimilation (A). However, measurement of leaf-level g sw response curves to changes in A are rare, particularly in the tropics, resulting in only limited data to evaluate this key assumption. Here, we measured the response of g sw and A to irradiance in six tropical species at different leaf phenological stages. We showed that the relationship between g sw and A was not linear, challenging the key assumption upon which optimality theory is based-that the marginal cost of water gain is constant. Our data showed that increasing A resulted in a small increase in g sw at low irradiance, but a much larger increase at high irradiance. We reformulated the popular Unified Stomatal Optimization (USO) model to account for this phenomenon and to enable consistent estimation of the key conductance parameters g 0 and g 1 . Our modification of the USO model improved the goodness-of-fit and reduced bias, enabling robust estimation of conductance parameters at any irradiance. In addition, our modification revealed previously undetectable relationships between the stomatal slope parameter g 1 and other leaf traits. We also observed nonlinear behavior between A and g sw in independent datasets that included data collected from attached and detached leaves, and from plants grown at elevated CO 2 concentration. We propose that this empirical modification of the USO model can improve the measurement of g sw parameters and the estimation of plant and ecosystem-scale water and CO 2 fluxes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Assessing the limitations of commercial sensors and models for supporting marine carbon dioxide removal monitoring: a case study

Several unknowns remain surrounding marine Carbon Dioxide Removal (mCDR) monitoring, reporting, and verification (MRV) practices and capabilities. Current in-situ sensor technology is limited (primarily pH and pCO 2 ), requiring calculations and assumptions to estimate changes in carbonate chemistry parameters, including total alkalinity (TA). Considering that cost, energy consumption, and accuracy of commercial sensors can vary by orders of magnitude, understanding how well existing sensors perform in an mCDR context is important for this emerging community. Likewise, documenting sensor limitations and how relatively simple models can optimize sensor deployments will improve MRV efforts and support protocol development. Here we (1) compare performance a variety of commercially available sensors in a blind mesocosm experiment simulating ocean alkalinity enhancement (OAE), and how sensor performance impacted carbonate chemistry estimates; (2) evaluate if sensors can distinguish the OAE signal from natural variability during a small scale OAE field test in Sequim Bay, WA, USA, and (3) use an idealized ocean biogeochemistry model to explore optimal sensor network design based on (1) and (2). Our mesocosm results indicate that correctly constraining pH uncertainty will be critical for accurate TA estimates with current sensor technology compared to the less impactful variation caused by uncertainty in pCO 2 (pH data that are presented throughout are reported on the total scale (pH T ) unless otherwise noted). Our pilot field test demonstrated that sensors were capable of distinguishing mCDR signatures from natural variability under optimal real-world conditions. Idealized modeling simulations of the field test showed that a range of sparse and dense (3 to 100) sensors sampling areas of detectable increases will underestimate the net change in surface pH by at least 35–55%, at both realistic and highly elevated alkalinity input levels. We also highlight the limitations of current sensing technology for MRV, and the importance of ocean biogeochemistry models as critical tools for predicting when and where mCDR signals will be detectable using available sensors. Overall, our findings suggest that commercially available pCO 2 sensors and some pH sensors will form an important backbone for mCDR MRV tasks, though complete MRV characterization will require these data to be used in combination with other tools.

OAE↗
DOE OSTI2021

A New Modeling Framework for Geothermal Operational Optimization with Machine Learning (GOOML)

Geothermal power plants are excellent resources for providing low carbon electricity generation with high reliability. However, many geothermal power plants could realize significant improvements in operational efficiency from the application of improved modeling software. Increased integration of digital twins into geothermal operations will not only enable engineers to better understand the complex interplay of components in larger systems but will also enable enhanced exploration of the operational space with the recent advances in artificial intelligence (AI) and machine learning (ML) tools. Such innovations in geothermal operational analysis have been deterred by several challenges, most notably, the challenge in applying idealized thermodynamic models to imperfect as-built systems with constant degradation of nominal performance. This paper presents GOOML: a new framework for Geothermal Operational Optimization with Machine Learning. By taking a hybrid data-driven thermodynamics approach, GOOML is able to accurately model the real-world performance characteristics of as-built geothermal systems. Further, GOOML can be readily integrated into the larger AI and ML ecosystem for true state-of-the-art optimization. This modeling framework has already been applied to several geothermal power plants and has provided reasonably accurate results in all cases. Therefore, we expect that the GOOML framework can be applied to any geothermal power plant around the world.

15 GEOTHERMAL ENERGY↗
DOE OSTI2023

Journey Over Destination: Differentiable Sensor Placement Enhances Generalization [Poster]

The challenge of reconstructing spatial fields that change over time from limited sensor data has been a focal point for many research studies. Various machine learning methods have been used in attempts to address this complex issue, including convolutional neural networks. All of the proposed methods share a common requirement that the user needs to manually determine the sensor positions. This requirement remains a limiting factor in the ongoing quest for efficient learning and accurate field reconstruction. This study aims to present a method that enables a model to optimize sensor positions via backpropagation, thereby facilitating the model’s exploration of the spatial domain and enhancing sensor positioning effectively. Indexing naturally incorporates discrete decisions. This operation is nondifferentiable which is a requirement for the application of gradient-based optimization methods. We showcased its effectiveness by training an attention-based neural network, which achieved top-tier performance on two separate datasets. To our knowledge, this represents the first fully end-to-end differentiable workflow for enhancing sensor placement within a neural network model.

58 GEOSCIENCES↗
DOE OSTI2025

Monitoring Sulfuric Acid and Temperature Using Raman Spectroscopy and Multivariate Chemometrics

Multivariate regression models were optimized for the quantification of sulfuric acid (H 2 SO 4 ) [0–8 M] and temperature (20 °C–80 °C) in the presence of ammonium sulfate ((NH 4 ) 2 SO 4 [0–0.6 M]) using Raman spectroscopy. Optical vibrational spectroscopy is a useful nondestructive technique for the in situ analysis of complex chemical systems notoriously difficult to monitor in situ and in real-time. Multivariate analysis, a chemometrics method, can be paired with these nondestructive optical methods for determining analyte concentration and speciation in complex solutions, such as dissociated species in polyprotic acids, e.g., H 2 SO 4 . The effect of temperature is often overlooked although it can have a major influence on speciation and the corresponding Raman spectra. Here, in this study, partial least squares regression models were optimized for the quantification of H 2 SO 4 and its two deprotonated forms as a function of temperature. Measuring bisulfate as a function of temperature is particularly challenging owing to changes in the second dissociation constant. A designed training set effectively minimized the sample set size and trained a robust predictive model with percent root mean square error of <3% for H 2 SO 4 . The practical strategy employed here was demonstrated to be effective for building chemometric models that directly account for dynamic temperatures with static samples and is shown to be amenable to flow cell analysis applications with a simple calibration transfer for process monitoring applications.

D-optimal design↗
DOE OSTI2021

A Comparative Study of Model Predictive Control and Optimal Causal Control for Heaving Point Absorbers

Efforts by various researchers in recent years to design simple causal control laws that can be applied to WEC devices suggest that these controllers can yield similar levels of energy output as those of more complex non-causal controllers. However, most studies were established without adequately considering device and power conversion system constraints which are relevant design drivers from a cost and economic point of view. It is therefore imperative to understand the benefits of MPC compared to causal control from a performance and constraint handling perspective. In this paper, we compare linear MPC to a casual controller that incorporates constraint handling to benchmark its performance on a one DoF heaving point absorber in a range of wave conditions. Our analysis demonstrates that MPC provides significant performance advantages compared to an optimized causal controller, particularly if significant constraints on device motion and/or forces are imposed. We further demonstrate that distinct control performance regions can be established that correlate well with classical point absorber and volumetric limits of the wave energy conversion device.

42 ENGINEERING↗