SEARCH · Engineering Papers

Results for “modeling and 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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 541 records · Page 30

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Moltensaltpropnet

MoltenSaltPropnet is a physics-informed machine learning framework that aims to predict the thermophysical properties of molten fluoride and chloride salt mixtures, which are crucial for the design and safety of Generation IV molten salt reactors. The code processes data from the Molten-Salt Thermal Properties Database (MSTDB-TP) and the Janz compendium, converting critically evaluated correlations into fast, differentiable surrogate models for density, viscosity, thermal conductivity, and heat capacity across 448 distinct salt systems. The implementation consists of several key components: 1. Data Curation: The code parses and cleans the raw data, normalizing elemental mole fractions and extracting relevant regression coefficients for various thermophysical properties. 2. Feature Engineering: It generates fixed-length numerical descriptors that encapsulate the composition and temperature, incorporating polynomial interaction terms and dimensionality-reduction techniques to optimize model performance. 3. Coefficient Learning: Four different machine learning architectures are employed: a deep residual network (ResNet), a Kolmogorov–Arnold network (KAN), a sparsity-inducing neural network (SNN), and classical regression models. Each model learns to predict coefficients that define the temperature-dependent correlations for the thermophysical properties. 4. Property Reconstruction: The predicted coefficients are used to compute temperature-dependent property values, ensuring positivity and monotonic trends through a composite loss function that enforces physical constraints. 5. User Interface: An open-source web application enables users to filter the database, train task-specific models, and visualize the results, allowing for rapid exploration of candidate salt mixtures. MoltenSaltPropnet bridges the gap between limited experimental data and high-fidelity reactor simulations, providing a powerful tool for researchers in the field of molten salt reactors and advanced nuclear energy systems.

Retamales, Mauricio Eduardo Tano [Idaho National L↗

DOE OSTI2020

Reaction Mechanism of Vapor-Phase Formic Acid Decomposition over Platinum Catalysts: DFT, Reaction Kinetics Experiments, and Microkinetic Modeling

A combination of periodic density functional theory (DFT, PW91-GGA) calculations, reaction kinetics experiments, and mean-field microkinetic modeling is used to derive insights on the reaction mechanism and determine the nature of the active site under reaction conditions for the vapor-phase decomposition of formic acid (FA, HCOOH) over Pt/C catalysts. Microkinetic models formulated using DFT energetics derived on the clean Pt(100) and Pt(111) required large parameter adjustments to reproduce the experimentally measured apparent activation energies and reaction orders. Further, these models predicted high surface coverage of adsorbed carbon monoxide (CO*), inconsistent with the environment of the active site in the DFT calculations on the clean surfaces. Consequently, we reperformed DFT calculations for the entire reaction network on partially CO*-covered (4/9 monolayer, ML) Pt(111) and Pt(100). The resultant microkinetic models, with thermochemistry and kinetics explicitly dependent on CO* coverage, were able to reproduce the experimentally determined activation energies and reaction orders, in addition to being self-consistent in CO* coverage. Our results suggest that Pt(100) is likely poisoned by CO* under typical reaction conditions and does not contribute significantly to the experimentally observed reactivity. Instead, we find that Pt(111) better represents the active site for FA decomposition reaction on Pt/C catalysts. The optimized model on 4/9 ML CO*-covered Pt(111) suggests that the reaction occurs via the carboxyl (COOH*) intermediate and that the spectator CO*-assisted pathways play a significant role under reaction conditions. Lastly, this study underscores the importance of spectator species on the energetics and the mechanism of a catalytic reaction and their key role in developing a model that better addresses the nature of the active site under realistic catalytic reaction conditions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Systems Analysis Approach to Polyethylene Terephthalate and Olefin Plastics Supply Chains in the Circular Economy: A Review of Data Sets and Models

The environmental and economic impacts of implementing a circular economy in plastic waste supply chains are not well understood. The proposed systems analysis framework assesses environmental, social, and economic impacts of plastic waste supply chains in a circular economy. The first objective of this article is to identify datasets, models and knowledge gaps associated with waste plastic supply chain processes, mainly in the U.S. Our literature review indicated that the best datasets exist for virgin plastic resin production, mechanical recycling, landfilling, and incineration, with materials recovery facility being intermediate, and with chemical recycling the lowest. The second objective of this article is to develop an illustrative application of the framework by conducting a preliminary systems analysis of PET bottles with closed-loop recycling. Here, the preliminary systems analysis of PET bottles utilized a linear programming optimization method. Our optimization model indicated that both chemical and mechanical recycling processes are needed to achieve a true circular economy of PET bottles with the least greenhouse gas emissions, specifically reductions of 24% when compared with the linear economy. Good quality and standardized life cycle assessment and techno-economic analysis studies are needed to better understand the environmental, economic, and social impacts of advanced sorting and chemical recycling technologies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Magnetic excitations of the hybrid multiferroic (ND 4 ) 2 FeCl 5 ·D 2 O

We report a comprehensive inelastic neutron scattering study of the hybrid molecule-based multiferroic compound ( ND 4 ) 2 FeCl 5 · D 2 O in the zero-field incommensurate cycloidal phase and the high-field quasicollinear phase. The spontaneous electric polarization changes its direction concurrently with the field-induced magnetic transition, from mostly aligned with the crystallographic a axis to the c axis. To account for such a change in polarization direction, the underlying multiferroic mechanism was proposed to switch from the spin-current model induced via the inverse Dzyaloshinskii-Moriya interaction to the p - d hybridization model. We perform a detailed analysis of the inelastic neutron data of ( ND 4 ) 2 FeCl 5 · D 2 O using linear spin-wave theory to quantify magnetic interaction strengths and investigate the possible impact of different multiferroic mechanisms on the magnetic couplings. Overall our result reveals that the spin dynamics of both multiferroic phases can be well described by a Heisenberg Hamiltonian with easy-plane anisotropy. We do not find notable differences between the optimal model parameters of the two phases. The hierarchy of exchange couplings and the balance among frustrated interactions remain the same between two phases, suggesting that magnetic interactions in ( ND 4 ) 2 FeCl 5 · D 2 O are much more robust than the electric polarization in response to delicate reorganizations of the electronic degrees of freedom in an applied magnetic field.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

DOE OSTI2023

Nuclear-Renewable-Storage Systems: Enhancing Planning and Operations of Integrated Energy Systems

Nuclear-renewable-storage integrated energy systems (IES) are multi-carrier energy systems that include not only electricity but also other forms of demands. Because individual IES components must observe their thermo-physical limits, including ramp rates, start-up, and shut-down time, we formulate operations of the IES as an optimization model by minimizing the total operations costs subject to physical limits of all constituent components. In addition, we develop a data-driven approach to improve the computational performance of the economic dispatch model by using reinforcement learning, where an agent is rewarded by meeting demands and penalized otherwise when shifting to the next state.

25 ENERGY STORAGE↗

NASA NTRS1993

Extended ecosystem signatures with application to Eos synergism requirements

The primary objective is to define the advantages of synergistically combining optical and microwave remote sensing measurements for the determination of biophysical properties important in ecosystem modeling. This objective was approached in a stepwise fashion starting with ground-based observations of controlled agricultural and orchard canopies and progressing to airborne observations of more natural forest ecosystems. This observational program is complemented by a parallel effort to model the visible reflectance and microwave scattering properties of composite vegetation canopies. The goals of the modeling studies are to verify our basic understanding of the sensor-scene interaction physics and to provide the basis for development of inverse models optimized for retrieval of key biophysical properties. These retrieval algorithms can then be used to simulate the expected performance of various aspects of Eos including the need for simultaneous SAR and HIRIS observations or justification for other (non-synchronous) relative timing constraints and the frequency, polarization, and angle of incidence requirements for accurate biophysical parameter extractions. This program completed a very successful series of truck-mounted experiments, made remarkable progress in development and validation of optical reflectance and microwave scattering models for vegetation, extended the scattering models to accommodate discontinuous and periodic canopies, developed inversion approaches for surface and canopy properties, and disseminated these results widely through symposia and journal publications. In addition, the third generation of the computer code for the microwave scattering models was provided to a number of other US, Canadian, Australian, and European investigators who are currently presenting and publishing results using the MIMICS research code.

Ulaby, Fawwaz T.↗

DOE OSTI2020

Rapid Identification of X-ray Diffraction Patterns Based on Very Limited Data by Interpretable Convolutional Neural Networks

Large volumes of data from material characterizations call for rapid and automatic data analysis to accelerate materials discovery. Herein, we report a convolutional neural network (CNN) that was trained based on theoretical data and very limited experimental data for fast identification of experimental X-ray diffraction (XRD) patterns of metal–organic frameworks (MOFs). To augment the data for training the model, noise was extracted from experimental data and shuffled; then it was merged with the main peaks that were extracted from theoretical spectra to synthesize new spectra. For the first time, one-to-one material identification was achieved. Theoretical MOFs patterns (1012) were augmented to a whole data set of 72 864 samples. It was then randomly shuffled and split into training (58 292 samples) and validation (14 572 samples) data sets at a ratio of 4:1. For the task of discriminating, the optimized model showed the highest identification accuracy of 96.7% for the top 5 ranking on a test data set of 30 hold-out samples. Neighborhood component analysis (NCA) on the experimental XRD samples shows that the samples from the same material are clustered in groups in the NCA map. Analysis on the class activation maps of the last CNN layer further discloses the mechanism by which the CNN model successfully identifies individual MOFs from the XRD patterns. Furthermore, this CNN model trained by the data augmentation technique would not only open numerous potential applications for identifying XRD patterns for different materials, but also pave avenues to autonomously analyze data by other characterization tools such as FTIR, Raman, and NMR spectroscopies.

36 MATERIALS SCIENCE↗

NASA NTRS1995

Analysis of the Hessian for aeroelastic optimization

The symbol of the Hessian for an aeroelastic optimization model problem is analyzed. The flow is modeled by the small-disturbance full potential equation and the structure is modeled by an isotropic (von Karman) plate equation. The cost function consists of both aerodynamic and structural terms. In the new analysis the symbol of the cost function Hessian near the minimum is computed. The result indicates that under some conditions, which are likely fulfilled in most applications, the system is decoupled for the non-smooth components. The result also shows that the structure part in the Hessian is well-conditioned while the aerodynamic part is ill-conditioned. Applications of the result to optimization strategies are discussed.

Arian, Eyal↗

DOE OSTI2020

Quantitative Understanding and Implementation of Screen Printed p+ Poly-Si/Oxide Passivated Contact to Enhance the Efficiency of p-PERC Cells

This paper reports on the modeling, optimization and implementation of p-TOPCon (tunnel oxide passivated contacts) on the rear side of a PERC to enhance its efficiency. Local Al-BSF of a traditional PERC was replaced by p+ polySi/oxide passivated contact composed of ~15Å thick chemically grown tunnel oxide, capped with 120-250nm thick p+ poly-Si layer grown by LPCVD. Process optimization resulted in full area unmetallized saturated current density (Jo) of ~ 5fA/cm2 for planar surface, nearly independent of poly-Si thickness in the range. Metallized J0 showed an increase with decreased poly-Si thickness and was found to be 9.6 and ~25fA/cm2 for 250nm and 120 nm polySi, respectively, with 4.6% direct metal-Si contact fraction, suitable for bifacial cells. A 21.4% efficient baseline PERC cell with local BSF was fabricated and analyzed to extract the rear side saturation current density (J0b’) of 66fA/cm2. Model calculations showed that by replacing this local BSF with 250nm TopCon developed in the study showed some Voc enhancement of 7mV, consistent with the observed Voc increase of 10mV. Model calculations also reveal that a more advanced LBSF PERC with better bulk lifetime and emitter saturation current density can extends its potential gain up to 0.4% in cell efficiency from the integration of p-TOPCon.

14 SOLAR ENERGY↗

DOE OSTI2023

Accelerating crystal structure determination with iterative AlphaFold prediction

Experimental structure determination can be accelerated with artificial intelligence (AI)-based structure-prediction methods such as AlphaFold . Here, an automatic procedure requiring only sequence information and crystallographic data is presented that uses AlphaFold predictions to produce an electron-density map and a structural model. Iterating through cycles of structure prediction is a key element of this procedure: a predicted model rebuilt in one cycle is used as a template for prediction in the next cycle. This procedure was applied to X-ray data for 215 structures released by the Protein Data Bank in a recent six-month period. In 87% of cases our procedure yielded a model with at least 50% of C α atoms matching those in the deposited models within 2 Å. Predictions from the iterative template-guided prediction procedure were more accurate than those obtained without templates. It is concluded that AlphaFold predictions obtained based on sequence information alone are usually accurate enough to solve the crystallographic phase problem with molecular replacement, and a general strategy for macromolecular structure determination that includes AI-based prediction both as a starting point and as a method of model optimization is suggested.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Boosting Noise2Inverse via enhanced model selection for denoising computed tomography data

Synchrotron-based x-ray tomographic imaging enables the examination of the internal structure of materials at high spatial and temporal resolution. Experimental constraints can impose dose and time limits on the measurements, introducing a higher level of noise and artifacts in the reconstructed images. Deep learning has emerged as a powerful tool to remove noise from reconstructed images. Recently, the Noise2Inverse method was designed specifically for denoising reconstructed images without requiring paired noisy and clean images. This method creates multiple statistically independent reconstructions used to pair the data in which training involves transforming one reconstruction into the other, and vice versa. Originally designed to be used after a fixed number of epochs, we see in practice that this approach may not produce the optimal model and may unnecessarily waste computational resources. Therefore, we propose an alternative method of identifying the best model during training that aligns with the Noise2Inverse method. During validation, we compare the model output of the multiple reconstructions among each other. We hypothesize that the best model is the one that produces images with the highest similarity, implying a convergence in the predicted material properties and absorption values. To compare model outputs, we consider the absolute error, square error, structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and cosine similarity. We evaluate our method on two simulated tomography datasets and two, real-world, low-contrast, high-energy x-ray tomography datasets. We show our approach is more effective at determining the best model, up to an increase of 12.50% and 12.53% in SSIM and PSNR, respectively, while only requiring a fifth of the training time compared to the original approach.

CT↗

DOE OSTI2021

Real-Time Dispatch With Secondary Frequency Regulation: A Pathway to Consider Intra-Interval Fluctuations

Real-time dispatch balances the power demand with minimized operating costs. For the current dispatch model, the demand is assumed to be constant within a time interval, while the intra-interval power balance is left to frequency regulation. Based on practical experience and simulations, this behavior may lead to insufficient frequency regulation and uneconomic regulation costs considering the increase in intra-interval fluctuations caused by renewables. Here, a real-time dispatch method with secondary frequency regulation behaviors is proposed. Without changing the interval of the real-time dispatch command, the system regulation mileage and intra-interval generation adjustment are explicitly formulated in a mixed-integer optimization model. To reduce the computational burden, an efficient two-stage calculation method is proposed. With practical utility data, this article finds that the proposed method can effectively improve the system frequency performance with the subminute net load forecasting curve obtained by an off-the-shelf forecasting approach. Time-domain simulations in IEEE and practical utility systems validate the effectiveness of the proposed method in terms of the frequency performance and total operating costs.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2022

Optimal Control of Biomass Feedstock Processing System Under Uncertainty in Biomass Quality

Planning of biorefinery operations is complicated by the stochastic nature of physical and chemical characteristics of biomass feedstock, such as, moisture level and carbohydrate content. Biomass characteristics affect the performance of the equipment which feed the reactor and the efficiency of the conversion process in a biorefinery. We propose a stochastic optimization model to identify a blend of feedstocks, inventory levels, and operating conditions of equipment to ensure a continuous flowing of biomass to the reactor while meeting the requirements of the biochemical conversion process. We propose a sample average approximation (SAA) of the model, and develop an efficient algorithm to solve the SAA model. A feedstock preprocessing process consists of two-stage grinding and pelleting is used to develop a case study. Extensive numerical analysis are conducted which lead to a number of observations. Our main observation is that sequencing bales based on moisture level and carbohydrate content leads to robust solutions that improve processing time and processing rate of the reactor. We provide a number of managerial insights that facilitate the implementation of the model proposed. Note to Practitioners—This paper is motivated by the challenges faced in the bioenergy industry. The focus of this paper is on plants which use the biochemical conversion process to generate liquid fuels. It has been observed that variations in biomass characteristics, such as moisture content, cause variations in feeding of the system which lead to under-utilization of equipment. A requirement of biochemical conversion process is to maintain the carbohydrate content of biomass processed by the reactor, larger than a threshold. We propose a model that identifies the inventory levels and operating conditions of equipment to ensure a continuous flowing of biomass to the reactor. The goal is to improve equipment utilization while satisfying the requirements of the conversion process. The model is tested using real-life data. We found out that by sequencing bales based on moisture level and carbohydrate content, a plant can reduce variability in the system leading to improved system reliability, higher processing rates of the reactor, and higher throughput.

09 BIOMASS FUELS↗

DOE OSTI2023

Hierarchical Flexibility Offering Strategy for Integrated Hybrid Resources in Real-time Energy Markets

This paper proposes a hierarchical model for determining the energy flexibility offering strategy of integrated hybrid resources (IHRs) in power distribution systems to participate in real-time energy markets. The proposed model utilizes the scalability, fast response time, and uncertainty observation of deep reinforcement learning (DRL) to overcome the scalability issue of operating numerous flexible resources and deliverability of energy flexibility to the real-time markets in the presence of the network constraints. To that end, the power distribution system is divided into multiple IHRs, where different types of flexible loads, energy storage systems, and solar plants with controllable inverters are operated through local IHR controllers, trained by deep deterministic policy gradient (DDPG) algorithm. Active power request and reactive power capacity of IHRs are then transmitted to a central flexibility controller, where a quadratic optimization model ensures the deliverability of the energy flexibility to the real-time energy market by satisfying the distribution network constraints. The proposed model is implemented on the 123-bus test power distribution system, demonstrating the capability of DRL-based hierarchical model for scalable operation of IHRs in order to offer deliverable energy flexibility to the real-time energy market.

Majidi, Majid↗

NASA NTRS2009

Massively Parallel Dantzig-Wolfe Decomposition Applied to Traffic Flow Scheduling

Optimal scheduling of air traffic over the entire National Airspace System is a computationally difficult task. To speed computation, Dantzig-Wolfe decomposition is applied to a known linear integer programming approach for assigning delays to flights. The optimization model is proven to have the block-angular structure necessary for Dantzig-Wolfe decomposition. The subproblems for this decomposition are solved in parallel via independent computation threads. Experimental evidence suggests that as the number of subproblems/threads increases (and their respective sizes decrease), the solution quality, convergence, and runtime improve. A demonstration of this is provided by using one flight per subproblem, which is the finest possible decomposition. This results in thousands of subproblems and associated computation threads. This massively parallel approach is compared to one with few threads and to standard (non-decomposed) approaches in terms of solution quality and runtime. Since this method generally provides a non-integral (relaxed) solution to the original optimization problem, two heuristics are developed to generate an integral solution. Dantzig-Wolfe followed by these heuristics can provide a near-optimal (sometimes optimal) solution to the original problem hundreds of times faster than standard (non-decomposed) approaches. In addition, when massive decomposition is employed, the solution is shown to be more likely integral, which obviates the need for an integerization step. These results indicate that nationwide, real-time, high fidelity, optimal traffic flow scheduling is achievable for (at least) 3 hour planning horizons.

Rios, Joseph Lucio↗

NASA NTRS1994

Asymptotic Giant Branch stars as a source of short-lived radioactive nuclei in the solar nebula

We carried out a theoretical evaluation of the contribution of Asymptotic Giant Branch (AGB) stars to some short-lived (10(exp 6) less than or equal to Tau-bar less than or equal to 2 x 10(exp 7) yr) isotopes in the Interstellar Medium (ISM) and in the early solar system using stellar model calculations for thermally pulsing evolutionary phases of low-mass stars. The yields of s-process nuclei in the convective He-shell for different neutron exposures tau(sub 0) were obtained, and AGB stars were shown to produce several radioactive nuclei (especially Pd-107, Pb-205, Fe-60, Zr-93, Tc-99, Cs-135, and Hf-182) in diferent amounts. Assuming either contamination of the solar nebula from a single AGB star or models for continuous injection and mixing from many stars into the ISM, we calculate the ratios of radioactive to stable nuclei at the epoch of the Sun's formation. The dilution factor between the AGB ejecta and the early solar system matter is obtained by matching the observed Pd-107/Pd-108 and depends on the value of tau(sub 0). It is found that small masses M(sub He) of He-shell material (10(exp -4)-10(exp -7) solar mass) enriched in s-process nuclei are sufficient to contaminate 1 solar mass of the ISM to produce the Pd-107 found in the early solar system. Predictions are made for all of the other radioactive isotopes. The optimal model to explain several observed radioactive species at different states of the proto-solar nebula involves a single AGB star with a low neutron exposure (tau(sub 0) = 0.03 mbarn(sup -1)) which contaminated the cloud with a dilution factor of M(sub He)/solar mass approximately 1.5 x 10(exp -4). This will also contribute newly synthesized stable s-process nuclei in the amount of approximately 10(exp -4) of their abundances already present in the proto-solar cloud. Variations in the degree of homogenization (approximately 30%) of the injected material may account for some of the small general isotopic anomalies found in meteorites. It is also found that Fe-60 is produced in small but significant quantities that may be sufficient to explain the observations if the time elapsed delta from the contamination of the ISM to the formation of protoplanetary bodies is not higher than delta = 5 x 10(exp 6) yr. If delta is longer, up to 10 x 10(exp 6) yr, this would require the single AGB star to experience enhanced neutron densities (n(sub n) approximately 3 x 10(exp 9)n/cu cm) in the s-processing zone in order to compensate for the branching at Fe-59. The alternative model of long-term continuous ejection of matter from many AGB stars does not appear to match the observations. We also estimate the Al-26 production from the H-shell and find that the Al-26 abundance in the early solar system may be readily explained in a self-consistent manner. Moreover, Al-26 from AGB stars may contribute substantially to the galactic Al-26 gamma-source, while no significant gamma-flux from Co-60 (deriving from Fe-60 decay) is to be expected.

Wasserburg, G. J.↗

DOE OSTI2025

Networked Microgrid Topology Reconfiguration to Promote Fairness in Proactive Load Shedding

Increasing occurrences of natural disasters and grid emergency events consistently challenge the safe and reliable operations of power systems. During such emergency situations, system operators may proactively shed load to mitigate risks. However, uncoordinated implementation of load shedding may disrupt electricity supply and even lead to cascading failures. Meanwhile, it is crucial to address potential biases affecting different customers when executing load shedding. This paper addresses the dynamic topology reconfiguration problem for networked microgrids with distributed energy resources under emergency conditions. Specifically, we propose a novel rolling-horizon optimization model that integrates fairness-aware constraints into the networked microgrid topology reconfiguration. Unlike existing approaches that focus solely on efficiency or apply fairness considerations in static settings, our method explicitly incorporates temporal fairness constraints to restrict repeated or excessive load curtailment for load blocks. Moreover, the fairness-aware constraints are specifically developed for the context of dynamic networked microgrid topology reconfiguration, and are designed to be convex or amenable to linear reformulations, which offers a more tractable alternative to traditional models with non-convex formulations. Numerical studies on a modified IEEE 13-bus system and a larger-sized SMART-DS networked microgrid system demonstrate the performance of the proposed algorithm towards more fairness-aware networked microgrid topology reconfiguration decision-making.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2022

Particle-based fast jet simulation at the LHC with variational autoencoders

Abstract We study how to use deep variational autoencoders (VAEs) for a fast simulation of jets of particles at the Large Hadron Collider. We represent jets as a list of constituents, characterized by their momenta. Starting from a simulation of the jet before detector effects, we train a deep VAE to return the corresponding list of constituents after detection. Doing so, we bypass both the time-consuming detector simulation and the collision reconstruction steps of a traditional processing chain, speeding up significantly the events generation workflow. Through model optimization and hyperparameter tuning, we achieve state-of-the-art precision on the jet four-momentum, while providing an accurate description of the constituents momenta, and an inference time comparable to that of a rule-based fast simulation.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗