Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model Inference”

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 1,693 records · Page 94
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Discriminative versus generative approaches to simulation-based inference

Most of the fundamental, emergent, and phenomenological parameters of particle and nuclear physics are determined through parametric template fits. Simulations are used to populate histograms which are then matched to data. This approach is inherently lossy, since histograms are binned and low-dimensional. Deep learning has enabled unbinned and high-dimensional parameter estimation through neural likelihood(-ratio) estimation. We compare two approaches for neural simulation-based inference (NSBI): one based on discriminative learning (classification) and one based on generative modeling. These two approaches are directly evaluated on the same datasets, with a similar level of hyperparameter optimization in both cases. In addition to a Gaussian dataset, we study NSBI using a Higgs boson dataset from the FAIR Universe Challenge. We find that both the direct likelihood and likelihood ratio estimation are able to effectively extract parameters with reasonable uncertainties. For the numerical examples and within the set of hyperparameters studied, we found that the likelihood ratio method is more accurate and/or precise. Both methods have a significant spread from the network training and would require ensembling or other mitigation strategies in practice.

high energy physics↗
DOE OSTI2018

Alpha Air Sample Counting Efficiency Versus Dust Loading: Evaluation of a Large Data Set

Dust loading on air sample filters is known to cause a loss of efficiency for direct counting of alpha activity on the filters, but the amount of dust loading and the correction factor needed to account for attenuated alpha particles is difficult to assess. In this paper, correction factors are developed by statistical analysis of a large database of air sample results for a uranium and plutonium processing facility at the Savannah River Site. As is typically the case, dust-loading data is not directly available, but sample volume is found to be a reasonable proxy measure; the amount of dust loading is inferred by a combination of the derived correction factors and a Monte Carlo model. The technique compares the distribution of activity ratios [beta/(beta + alpha)] by volume and applies a range of correction factors on the raw alpha count rate. The best-fit results with this method are compared with MCNP modeling of activity uniformly deposited in the dust and analytical laboratory results of digested filters. Finally, a linear fit is proposed to evenly-deposited alpha activity collected on filters with dust loading over a range of about 2 mg cm -2 to 1,000 mg cm -2 .

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2026

Modeling the Cosmological Lyman-𝛼 Forest at the Field Level

The distribution of absorption lines in the spectra of distant quasars, called the Lyman-𝛼 (Ly-𝛼) forest, is a unique probe of cosmology and the intergalactic medium at high redshifts and small scales. The statistical power of ongoing redshift surveys demands precise theoretical tools to model the Ly-𝛼 forest. We address this challenge by developing an analytic, perturbative forward model to predict the Ly-𝛼 forest at the field level for a given set of cosmological initial conditions. Our model shows a remarkable performance when compared with the Sherwood hydrodynamic simulations: it reproduces the Ly-𝛼 forest flux power spectrum, its cross-correlation with dark matter halos, and the one-point probability distribution function of both fields at the percent level down to scales of a few Mpc. Our work provides crucial tools that bridge analytic modeling on large scales with simulations on small scales, enabling field-level inference from Ly-𝛼 forest data and simulation-based priors for cosmological analyses. Furthermore, this is especially timely for realizing the full scientific potential of the Ly-𝛼 forest measurements by the dark energy spectroscopic instrument.

Cosmological parameters↗
DOE OSTI2022

Symmetry of Cosmological Observables, a Mirror World Dark Sector, and the Hubble Constant

We find that a uniform scaling of the gravitational free-fall rates and photon-electron scattering rate leaves most dimensionless cosmological observables nearly invariant. This result opens up a new approach to reconciling cosmic microwave background and large-scale structure observations with high values of the Hubble constant H0: Find a cosmological model in which the scaling transformation can be realized without violating any measurements of quantities not protected by the symmetry. A “mirror world” dark sector allows for effective scaling of the gravitational free-fall rates while respecting the measured mean photon density today. Further model building might bring consistency with the two constraints not yet satisfied: the inferred primordial abundances of deuterium and helium.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2017

Electron Temperature Fluctuation Measurements and Transport Model Validation at Alcator C-Mod

The tokamak is a type of toroidal device used to confine a fusion plasma using large magnetic fields. Tokamaks and stellarators the leading devices for confining plasmas for fusion, and the capability to predict performance in these magnetically confined plasmas is essential for developing a sustainable fusion energy source. The magnetic configuration of tokamaks and stellarators does not exist in Nature, yet, the fundamental processes governing transport in fusion plasmas are universal – turbulence and instabilities, driven by inhomogeneity and asymmetry in the plasma, conspire to transport heat and particles across magnetic field lines and can play critical roles in impurity confinement and generation of intrinsic rotation. Turbulence exists in all plasmas, and in neutral fluids as well. The study of turbulence is essential to developing a fundamental understanding of the nature of the fourth state of matter, plasmas. Experimental studies of turbulence in tokamaks date back to early scattering observations from the late 1970s. Since that time, great advances in turbulence diagnostics have been made, all of which have significantly enhanced our knowledge and understanding of turbulence in tokamaks. Through comparisons with advanced gyrokinetic theory and turbulent-transport models a great deal of evidence exists to implicate turbulent-driven transport as an important mechanism determining transport in all channels: heat, particle and momentum However, prediction and control of turbulent-driven transport remains elusive. Key to development of predictive transport models for magnetically confined fusion plasmas is validation of the nonlinear gyrokinetic transport model, which describes transport due to turbulence. Validation of gyrokinetic codes must include detailed and quantitative comparisons with measured turbulence characteristics, in addition to comparisons with inferred transport levels and equilibrium profiles. For this reason, advanced plasma diagnostics for studying core turbulence are needed in order to assess the accuracy of gyrokinetic models for turbulent-driven particle, heat and momentum transport. New core turbulence diagnostics at the world-class tokamaks Alcator C-Mod at MIT and ASDEX Upgrade at the Max Planck Institute for Plasma Physics have been designed, developed, and operated over the course of this project. These new instruments are capable of measuring electron temperature fluctuations and the phase angle between density and temperature fluctuations locally and quantitatively. These new data sets from Alcator C-Mod and ASDEX Upgrade are being used to fill key gaps in our understanding of turbulent transport in tokamaks. In particular, this project has results in new results on the topics of the Transport Shortfall, the role of ETG turbulence in tokamak plasmas, profile stiffness, the LOC/SOC transition, and intrinsic rotation reversals. These data are used in a rigorous process of “Transport model validation”, and this group is a world-leader on using turbulence models to design new hardware and new experiments at tokamaks. A correlation electron cyclotron emission (CECE) diagnostic is an instrument used to measure micro-scale fluctuations (mm-scale, compared to the machine size of meters) of electron temperature in magnetically confined fusion plasmas, such as those in tokamaks and stellarators. These micro-scale fluctuations are associated with drift-wave type turbulence, which leads to enhanced cooling and mixing of particles in fusion plasmas and limits achieving the required temperatures and densities for self-sustained fusion reactions. A CECE system can also be coupled with a reflectometer system that measured micro-scale density fluctuations, and from these simultaneous measurements, one can extract the phase between the density (n) and temperature (T) fluctuations, creating an nT phase diagnostic. Measurements of the fluctuations and the phase angle between them are extremely useful for testing and validating predictive models for the transport of heat and particles in fusion plasmas due to turbulence. Once validated, the models are used to predict performance in ITER and other burning plasmas, such as the MIT ARC design. Most recently, data from the newly developed, so-called “CECE diagnostic” [Cima 1995, White 2008] and “nT phase angle measurements” [Haese 1999, White 2010] ]will be combined with data from density fluctuation diagnostics at ASDEX Upgrade to support a long-term program of physics research in turbulence and transport that will allow for more stringent testing and validation of gyrokinetic turbulent-transport codes. This work directly impacts the development of predictive transport models in the U.S. FES program, such as TGLF, developed by General Atomics, which are used to predict performance in ITER and other burning plasma devices as part of advancing the development of fusion energy sciences.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2017

UQTk Version 3.0.3 User Manual

The UQ Toolkit (UQTk) is a collection of libraries and tools for the quantification of uncertainty in numerical model predictions. Version 3.0.3 offers intrusive and non-intrusive methods for propagating input uncertainties through computational models, tools for sen- sitivity analysis, methods for sparse surrogate construction, and Bayesian inference tools for inferring parameters from experimental data. This manual discusses the download and installation process for UQTk, provides pointers to the UQ methods used in the toolkit, and describes some of the examples provided with the toolkit.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2018

A Scaling Study of Microbially-Enhanced Methane Production From Coal (Mecbm): Optimizing Nutrient Delivery For Maximized Methane Production

Microbially-enhanced coalbed methane (MECBM) recovery is a novel method to stimulate gas production in CBM reservoirs. We first develop suitable nutrient recipes for native coals from sites each in the Illinois Basin and in the San Juan Basin. The optimal concentrations for statistically significant parameters are identified and the methane yields at laboratory scale are obtained (Part I & II). We then demonstrate the importance of replicating truly in-situ conditions for bio- gasification studies by investigating methane productivity in a core holder simulating in-situ conditions. Based on the experimental results, the resulting gas production correlates positively with the permeability of the samples. We examine the deliverability of engineered microbial solutions to promote coal bioconversion in coals and demonstrate that greater than milli-darcy permeability is required under in-situ stress conditions (Part III). This constraint is supplemented by imaging and a fractal-based characterization approach to evaluate changes in the physical properties of coal due to bio-gasification. High resolution SEM imaging is used to observe swelling of the coal matrix with sub-micron wide continuous pores observed to persist even after microbial treatment (IV). We explore the potential of generating, propping-open and sustaining the functionality of hydraulic fractures in coal using different fracturing fluids (Part V). The alteration of the nano-pore structure due to MECBM treatment is examined by using small angle X-ray scattering (SAXS) to define the in-situ porosity evolution as a proxy to infer permeability evolution (Part VI). Coupled hydro-mechanical and bio-gasification modeling is applied to determine the effectiveness of nutrient delivery by hydraulic fracturing and its field implementation. Compared with both the unstimulated and hydro-shear-dilated cases of the natural fracture network, hydraulic fracturing demonstrates a significant improvement to enhance nutrient delivery. A preliminary analysis of field-scale methane yield is completed with consideration of the combined influence of permeability evolution, solute transport and microbial growth (Part VII). Developing nutrient solutions for Illinois Basin and San Juan Basin (Task 2): To enhance methane release from coal through bio-gasification, suitable nutrient solutions are needed to stimulate microbial activities toward coal depolymerization and conversion to biogas. Specific for bituminous coal in the Illinois basin, a nutrient recipe that can be used to enhance coal bio- gasification in situ is developed. Based on our previous experience working with Illinois coal and the characteristics of the formation water and the indigenous microbial community, four parameters, Fe powder, methanol, ethanol, and trace minerals are optimized through using a Box-Behnken design. The optimal condition predicted by the models is: Fe-powder at 74 mM, methanol at 97.91 mM, ethanol at 100 mM, and trace minerals at 100%. Under these conditions, the predicted methane yield and content is 1,417.35 ft 3 /ton and 80.7%, respectively. These results are then validated by experimental studies. In addition, each added component is evaluated in terms of its contribution to methane generated. Specifically, the role of coal in the bio-gasification process is studied against two other solid materials. Experimental data from different solid matrices prove that coal is indeed degraded by the studied microbial community.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2017

HESS J1427−608: AN UNUSUAL HARD, UNBROKEN γ -RAY SPECTRUM IN A VERY WIDE ENERGY RANGE

We report the detection of a GeV γ -ray source that spatially overlaps and is thus very likely associated with the unidentified very high energy (VHE) γ -ray source HESS J1427−608 with the Pass 8 data recorded by the Fermi Large Area Telescope . The photon spectrum of this source is best described by a power law with an index of 1.85 ± 0.17 in the energy range of 3–500 GeV, and the measured flux connects smoothly with that of HESS J1427−608 at a few hundred gigaelectronvolts. This source shows no significant extension and time variation. The broadband GeV to TeV emission over four decades of energies can be well fitted by a single power-law function with an index of 2.0, without obvious indication of spectral cutoff toward high energies. Such a result implies that HESS J1427−608 may be a PeV particle accelerator. We discuss the possible nature of HESS J1427−608 according to the multiwavelength spectral fittings. Given the relatively large errors, either a leptonic or a hadronic model can explain the multiwavelength data from radio to VHE γ -rays. The inferred magnetic field strength is a few micro-Gauss, which is smaller than the typical values of supernova remnants (SNRs) and is consistent with some pulsar wind nebulae (PWNe). On the other hand, the flat γ -ray spectrum is slightly different from typical PWNe but is similar to that of some known SNRs.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2017

Centrally Concentrated X-Ray Radiation from an Extended Accreting Corona in Active Galactic Nuclei

The X-ray emission from bright active galactic nuclei (AGNs) is believed to originate in a hot corona lying above a cold, geometrically thin accretion disk. A highly concentrated corona located within ∼10 gravitational radii above the black hole is inferred from observations. Based on the accretion of interstellar medium/wind, a disk corona model has been proposed in which the corona is well coupled to the disk by radiation, thermal conduction, as well as by mass exchange. Such a model avoids artificial energy input to the corona and has been used to interpret the spectral features observed in AGN. In this work, it is shown that the bulk emission size of the corona is very small for the extended accretion flow in our model. More than 80% of the hard X-ray power is emitted from a small region confined within 10 Schwarzschild radii around a non-spinning black hole, which is expected to be even smaller accordingly for a spinning black hole. Here, the corona emission is more extended at higher Eddington ratios. The compactness parameter of the corona, l=(L/R)((σ{sub T})/(m{sub e}c{sup 3})), is shown to be in the range of 1–33 for Eddington ratios of 0.02–0.1. Combined with the electron temperature in the corona, this indicates that electron–positron pair production is not dominant in this regime. A positive relation between the compactness parameter and photon index is also predicted. By comparing the above model predictions with observational features, we find that the model is in agreement with observations.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Late-onset Circumstellar Medium Interactions are Rare: An Unbiased GALEX View of Type Ia Supernovae

Abstract Using ultraviolet (UV) light curves, we constrain the circumstellar environments of 1080 Type Ia supernovae (SNe Ia) within z < 0.5 from archival Galaxy Evolution Explorer (GALEX) observations. All SNe Ia are required to have pre- and post-explosion GALEX observations to ensure adequate subtraction of the host-galaxy flux. Using the late-time GALEX observations, we look for the UV excess expected from any interaction between the SN ejecta and circumstellar material (CSM). Four SNe Ia are detected near maximum light, and we compare the GALEX photometry to archival data. However, we find that none of our targets show convincing evidence of CSM interaction. A recent Hubble Space Telescope (HST) survey estimates that ∼6% of SNe Ia may interact with distant CSM, but statistical inferences are complicated by the small sample size and selection effects. By injecting model light curves into our data and then recovering them, we constrain a broad range of CSM interactions based on the CSM interaction start time and the maximum luminosity. Combining our GALEX nondetections with the HST results, we constrain occurrence of late-onset CSM interaction among SNe Ia with moderate CSM interaction, similar to that observed in PTF11kx, to f CSM ≲ 5.1% between 0 and 500 days after discovery and ≲2.7% between 500 and 1000 days after discovery at 90% confidence. For weaker CSM interactions similar to SN 2015cp, we obtain limits of ≲16% and ≲4.8%, respectively, for the same time ranges.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2018

Remote Aerial Inspection Vehicle Design and Evaluation - 18428

This work details the system integration of an aerial vehicle with advanced capabilities for precise inspection in confined, unstructured environments. High levels of uncertainty and varying environmental conditions are significant challenges that degrade the performance of robot subsystems, including state estimation and navigation. The methods presented in this paper outline three methodologies to minimize uncertainty and maintain fast, safe navigation in cluttered, enclosed environments: (i) a state estimation framework that performs sensor fusion of laser odometry and localization using an Unscented Kalman Filter (ii) Robust Experience-driven Predictive Control, an adaptive control strategy to bound state uncertainty and guarantee safety in the presence of model perturbations (iii) an adaptive teleoperation scheme to generate smooth control inputs and infer user intent for an enhanced teleoperation interface. These methodologies are evaluated individually and implemented on a micro-aerial vehicle for testing with varying levels of noise and state uncertainty representing the conditions within the target domain. (authors)

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Modeling of melting and erosion in high velocity microparticle impacts

High velocity collisions can cause melting, a situation involving elasticity, plasticity, and phase transformations in tandem, which is very challenging for mechanical modeling. Here, this paper advances the use of smoothed particle hydrodynamics (SPH) simulations for this complex situation, investigating stainless steel microspheres impacting a tin substrate. After extensive validation against a range of mechanical and geometrical measurements of experimental impacts, the model provides a detailed, quantitative view of melt volume production over a broad range of impact velocities. It also enables direct decomposition of the impact energy partitioning, including analysis of the energy absorbed by plasticity in the impactor, the system kinetic energy, and the adiabatic heating of the substrate in addition to the melting itself. Since these quantities are not available experimentally, SPH modeling substantially enhances the quantitative understanding of such impacts. Finally, the model speaks to the phase characteristics of the ejecta formed during impact, confirming the experimental inference that the ejected material likely contains a significant fraction of solid material.

Constitutive modeling↗
DOE OSTI2019

An Adaptive Bayesian Parameter Estimation of a Synchronous Generator Under Gross Errors

Polynomial-chaos-expansion-based surrogate models have recently been advocated in the literature for power system dynamic parameter estimation. Regarding the estimation of the uncertain generator parameters, a Bayesian inference framework has been proposed based on a polynomial-based reduced-order representation of the synchronous machines using assumed parameter values. Then, the non-Gaussian posterior probability distribution functions (pdfs) of these parameters are recovered through the stochastic sampling approach efficiently. However, facing very large parameter errors, the reliability of the surrogate model decreases, yielding biased estimation results. To overcome this problem, here we develop hierarchical Bayesian inference framework that processes the measurements provided by phasor measurement units, while making use of multifidelity surrogates together with the importance sampling method. The latter allows us to estimate in an efficient manner the posterior pdfs of the uncertain parameters through the normalized weights of the prior samples. To improve the accuracy of the posterior pdfs, an adaptive procedure is further adopted in the importance sampling for the gradual evolution of its proposal functions. The new proposals assist in fine-tuning the sample space and thereby help to construct surrogates with higher fidelity. Through an iterative process, this approach is able to estimate accurately and efficiently non-Gaussian posterior pdfs of the uncertain generator parameters subject to gross errors.

42 ENGINEERING↗
DOE OSTI2016

DISK–WIND CONNECTION DURING THE HEARTBEATS OF GRS 1915+105

Disk and wind signatures are seen in the soft state of Galactic black holes, while the jet is seen in the hard state. Here we study the disk–wind connection in the ρ class of variability in GRS 1915+105 using a joint NuSTAR – Chandra observation. The source shows 50 s limit cycle oscillations. By including new information provided by the reflection spectrum and using phase-resolved spectroscopy, we find that the change in the inner disk inferred from the blackbody emission is not matched by reflection measurements. The latter is almost constant, independent of the continuum model. The two radii are comparable only if the disk temperature color correction factor changes, an effect that could be due to the changing opacity of the disk caused by changes in metal abundances. The disk inclination is similar to that inferred from the jet axis, and oscillates by ∼10°. The simultaneous Chandra data show the presence of two wind components with velocities between 500 and 5000 km s{sup −1}, and possibly two more with velocities reaching 20,000 km s{sup −1} (∼0.06 c ). The column densities are ∼5 × 10{sup 22} cm{sup −2}. An upper limit to the wind response time of 2 s is measured, implying a launch radius of <6 × 10{sup 10} cm. The changes in wind velocity and absorbed flux require the geometry of the wind to change during the oscillations, constraining the wind to be launched from a distance of 290–1300 r{sub g} from the black hole. Both data sets support fundamental model predictions in which a bulge originates in the inner disk and moves outward as the instability progresses.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

Experimental Investigation of Two-Phase Flow in Consolidating and Intact Salt: Analysis of Capillary Pressure Measurements from Core Labs

Two-phase flow of brine and air through very low permeability porous media is the dominant fluid flow state in and around geologic repositories for nuclear waste disposal. Constitutive models developed for other subsurface media (shale, sandstone, limestone, etc.) have been adopted but never experimentally parameterized for geologic salt. This report summarizes results of scoping capillary pressure and brine saturation measurements on reconsolidated run-of-mine salt and intact geologic salt samples. These measurements were obtained using mercury injection tests and yielded values of threshold pressure, pore-size distribution, and mercury saturation. The brine capillary pressure curve was then inferred from the raw mercury capillary pressure data. This report also summarizes estimates of the constitutive model parameters from the brine saturation and capillary pressure data (Brooks and Corey, 1964; van Genuchten, 1980). The results suggests that the model of van Genuchten (1980) is better suited for brine-air-salt systems. Additional testing is needed for the low brine saturation range of the capillary pressure curve where the present suite of tests did not yield sufficient data.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2019

Bottomonium suppression at RHIC and LHC

The strong suppression of heavy quarkonia is a good indicator that one has generated a quark-gluon plasma (QGP) in an ultrarelativistic heavy ion collision. Recent advancements in first principles calculations of the heavy quark potential provide additional insight into this suppression and can be used to further pin down the in-medium properties of the QGP. Realistic 3+1d dissipative hydrodynamical models can be used to simulate the QGP and, with bottomonium as a probe, one can make inferences for the initial temperature of the QGP and other phenomenological parameters such as the shear viscosity to entropy density ratio. However, progressing to LHC sNN=5.02TeV Pb-Pb collisions, one expects regeneration to have an increasingly important impact on the suppression observables. In this proceedings, we present a brief overview to set up the model and then provide model results for bottomonium suppression and regeneration, for ultrarelativistic heavy-ion collision experiments at RHIC and LHC

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

Predicting Missing Regions in Charged Particle Tracks Using a Sparse 3D Convolutional Neural Network

The 2x2 Demonstrator is a prototype detector for the Deep Underground Neutrino Experiment (DUNE)'s Near Detector. Both the 2x2 Demonstrator and the Near Detector itself will have inactive regions wherein there is no sensitivity to charge deposition and light signals that arise from charged particle interactions with liquid argon. In the 2x2, these inactive regions are positioned in-between the active detector modules, which introduces the challenge of inferring what charge signals ought to look like in these regions. This study explores the use of a Sparse 3D Convolutional Neural Network (ConvNet) to infer missing regions in charged particle tracks. Hits corresponding to energy depositions are voxelized into a three-dimensional (3D) grid for each track. Inactive regions within the tracks are replaced with a dense, rectangular 3D grid of voxels, ensuring consistent step sizes in X, Y, and Z directions. Voxels in these dense regions are initialized with an energy value of -1, indicating nonphysical energy or charge. The model is trained to predict which voxels should activate as part of the track and which should not, with the goal of eventually inferring the missing charge or energy values in these voxels. Results indicate that the model accurately predicts track voxels within ±1 unit in X, Y, or Z directions and effectively identifies non-track voxels, despite some overprediction. The approach shows promise in prediction of missing track regions with some accuracy.

43 PARTICLE ACCELERATORS↗
DOE OSTI2024

Predicting Missing Regions in Charged Particle Tracks Using a Sparse 3D Convolutional Neural Network

The 2x2 Demonstrator is a prototype detector for the Deep Underground Neutrino Experiment (DUNE)'s Near Detector. Both the 2x2 Demonstrator and the Near Detector itself will have inactive regions wherein there is no sensitivity to charge deposition and light signals that arise from charged particle interactions with liquid argon. In the 2x2, these inactive regions are positioned in-between the active detector modules, which introduces the challenge of inferring what charge signals ought to look like in these regions. This study explores the use of a Sparse 3D Convolutional Neural Network (ConvNet) to infer missing regions in charged particle tracks. Hits corresponding to energy depositions are voxelized into a three-dimensional (3D) grid for each track. Inactive regions within the tracks are replaced with a dense, rectangular 3D grid of voxels, ensuring consistent step sizes in X, Y, and Z directions. Voxels in these dense regions are initialized with an energy value of -1, indicating nonphysical energy or charge. The model is trained to predict which voxels should activate as part of the track and which should not, with the goal of eventually inferring the missing charge or energy values in these voxels. Results indicate that the model accurately predicts track voxels within 1 unit in X, Y, or Z directions and effectively identifies non-track voxels, despite some overprediction. The approach shows promise in prediction of missing track regions with some accuracy.

Utaegbulam, Hilary↗