Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “weight loss”

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 235 records · Page 13
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

A Comparison of Outer Rotor Radial and Axial Flux Machines for Application in Electric Vehicles

Axial flux machines have attracted a lot of interest in the recent years as potential high torque low weight candidates for use as traction motors in electric vehicles (EV). This paper compares axial and radial flux machines for electric vehicle applications. An external rotor radial flux machine with a Halbach array surface permanent magnet rotor and concentrated windings is chosen as a baseline to compare with axial flux designs. Both axial and radial flux motors are sized to meet the EV same requirements. Multi-objective design optimization using differential evolution minimizing loss and volume is carried out for both types of machines. Hundreds of candidate designs for each type of machine are analyzed, pareto fronts are identified and compared. The potential advantages of axial flux machines are evaluated and quantified.

Rallabandi, Vandana↗
DOE OSTI2025

Rare species do not disproportionately contribute to phylogenetic diversity in a subalpine plant community

Abstract Premise Within plant communities, few species are abundant, and most are locally rare. Worldwide, 36% of plant species are exceedingly rare and often face high extinction risk. However, the community phylogenetic impact of the loss of rare plants is largely unknown in many systems. We address this gap by investigating how rare species contribute to phylogenetic diversity, considering multiple metrics of rarity and multiple elevations in a subalpine plant community. Methods We collected abundance data at three sites near the Rocky Mountain Biological Laboratory (Colorado, USA). We calculated each species' range size from public occurrence data. We calculated phylogenetic signal for abundance and range size, compared community phylogenetic metrics weighted by range size and abundance to unweighted metrics, and quantified the change in phylogenetic diversity when removing single species and groups of species ranked by rarity. Results We found phylogenetic signal for abundance, but not range size. There was no difference between rarity‐weighted and ‐unweighted phylogenetic diversity metrics. Finally, phylogenetic diversity did not decline more when we removed single rare species or groups of rare species than when we removed single common species and groups of common species. Conclusions We found that rare species, whether at low abundance or with a small range, do not disproportionately contribute to phylogenetic diversity in our subalpine plant community. These results were consistent across elevations. Instead, rare species might provide phylogenetic redundancy with common species. Deeper understanding of functional differentiation is needed to understand contributions of rare species to this system.

Plant Sciences↗
DOE OSTI2020

Additive manufacturing of two-phase lightweight, stiff and high damping carbon fiber reinforced polymer microlattices

Carbon fiber reinforced polymer (CFRP) composite is known for its high stiffness-to-weight ratio and hence is of great interest in several engineering fields such as aerospace, automotive, defense, etc. However, such a composite is not suitable for energy dissipation as failure occurs with very little or no plastic deformation. Herein, we present an extendable multi-material projection microstereolithography process capable of producing carbon-fiber-reinforced cellular materials that achieve simultaneously high specific stiffness and damping coefficient. Inspired by the upper bounds of stiffness-loss coefficient in a two-phase composite, we designed and additively manufactured CFRP microlattices with soft phases architected into selected stiff-phase struts. Our results, confirmed by experimental and analytical calculations, revealed that the damping performance can be significantly enhanced by the addition of only a small fraction of the soft phase. The presented design and additive manufacturing strategy allow for optimizing mutually exclusive properties. As a result, these CFRP microlattices achieved high specific stiffness comparable to commercial CFRP, technical ceramics, and composites, while being dissipative like elastomers.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Modeling Functional Organic Chemistry in Arctic Rivers: An Idealized Siberian System

Rivers of the Arctic will become ever more important for the global climate, since they carry a majority of continental dissolved organic carbon flux into the rapidly changing polar ocean. Aqueous organics comprise a wide array of functional groups, several of which are likely to impact coastal and open water biophysical properties. Light attenuation, interfacial films, aerosol formation, gas release and momentum exchange can all be cited. We performed Lagrangian kinetic modeling for the evolution of riverine organic chemistry as the molecules in question make their way from the highlands to Arctic outlets. Classes as diverse as the proteins, sugars, lipids, re-condensates, humics, bio-tracers and small volatiles are all included. Our reduced framework constitutes an idealized northward flow driving a major hydrological discharge rate and primarily representing the Russian Lena. Mountainous, high solute and tundra sources are all simulated, and they meet up at several points between soil and delta process reactors. Turnover rates are parameterized beginning with extrapolated coastal values imposed along a limited tributary network, with connections between different terrestrial sub-ecologies. Temporal variation of our total dissolved matter most closely resembles the observations when we focus on the restricted removal and low initial carbon loads, suggesting relatively slow transformation along the water course. Thus, channel combinations and mixing must play a dominant role. Nevertheless, microbial and photochemical losses help determine the final concentrations for most species. Chemical evolution is distinct for the various functionalities, with special contributions from pre- and post-reactivity in soil and delta waters. Several functions are combined linearly to represent the collective chromophoric dissolved matter, characterized here by its absorption. Tributaries carry the signature of lignin phenols to segregate tundra versus taiga sources, and special attention is paid to the early then marine behaviors of low molecular weight volatiles. Heteropolycondensates comprise the largest percentage of reactive carbon in our simulations due to recombination/accumulation, and they tend to be preeminent at the mouth. Outlet concentrations of individual structures such as amino acids and absorbers lie above threshold values for biophysical influence, on the monolayer and light attenuation. The extent of coastal spreading is examined through targeted regional box modeling, relying on salinity and color for calibration. In some cases, plumes reach the scale of peripheral arctic seas, and amplification is expected during upcoming decades. Conclusions are mapped from the Lena to other boreal discharges, and future research questions are outlined regarding the bonding type versus mass release as permafrost degrades. Dynamic aqueous organic coupling is recommended for polar system models, from headwaters to coastal diluent.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Joule heating in bad and slow metals

Heat supplied to a metal is absorbed by the electrons and then transferred to the lattice. In conventional metals energy is released to the lattice by phonons emitted from the Lindhard continuum. However in a 'bad' metal, with short mean free path, the low energy Lindhard continuum is destroyed. To describe energy transfer to the lattice in these cases we obtain a general Kubo formula for the energy relaxation rate in terms of the electronic density spectral weight \text{lm} \, G^R_{nn}(\omega_{k},k) lm G n n R ( ω k , k ) evaluated on the phonon dispersion \omega_k ω k . We apply our Kubo formula to the high temperature Hubbard model, using recent data from quantum Monte Carlo and experiments in ultracold atoms to characterize \text{lm} \, G^R_{nn}(\omega_{k},k) lm G n n R ( ω k , k ) . We furthermore use recent data from electron energy-loss spectroscopy to estimate the energy relaxation rate of the cuprate strange metal to a high energy optical phonon. As a second, distinct, application of our formalism we consider 'slow' metals. These are defined to have Fermi velocity less than the sound velocity, so that particle-hole pairs are kinematically unable to emit phonons. We obtain an expression for the energy relaxation rate of a slow metal in terms of the optical conductivity.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Endpoint Use Efficiency Comparison for AC and DC Power Distribution in Commercial Buildings

Advances in power electronics and their use in Miscellaneous Electric Loads (MELs) in buildings have resulted in increased interest in using low-voltage direct current (DC) power distribution as a replacement for the standard alternating current (AC) power distribution in buildings. Both systems require an endpoint converter to convert the distribution system voltage to the MELs voltage requirements. This study focused on the efficiency of these endpoint converters by testing pairs of AC/DC and DC/DC power converters powering the same load profile. In contrast to prior studies, which estimated losses based on data sheet efficiency and rated loads, in this study, we used part load data derived from real-world time-series load measurements of MELs and experimentally characterized efficiency curves for all converters. The measurements performed for this study showed no systematic efficiency advantage for commercially available DC/DC endpoint converters relative to comparable, commercially available AC/DC endpoint converters. For the eight appliances analyzed with the pair of converters tested, in 50%, the weighted energy efficiency of the DC/DC converter was higher, while, for the other 50%, the AC/DC converter was. Additionally, the measurements indicated that the common assumption of using either data sheet efficiency values or efficiency at full load may result in substantial mis-estimates of the system efficiency.

AC/DC converters↗
DOE OSTI2020

Advanced Moderation Module for High-Temperature Micro-Reactor Applications

Development and deployment of micro-reactors that provide competitive efficiency, prevailing compactness, and inherent safety is an immediate focus of the U.S. nuclear industry. For thermal neutron micro-reactors operating at elevated temperature for optimized efficiency, such as micro molten-salt reactors (MSRs), heat-pipe reactors, and very-high temperature reactors (VHTRs), high-performance moderator based on metal hydride can enhance the neutron economy and therefore achieve reduced weight and enhanced portability. As unclad metal hydride inevitably decomposes at elevated temperature, an enclosure is required for hydride moderator to deliver desired performance at elevated temperatures. Conventional moderator enclosure solutions based on high-temperature alloys introduce extraneous neutron penalty into the micro-reactor, affecting the neutronic benefits provided by the hydride moderator. Additionally, the compatibility between the high-temperature alloy enclosure and the micro-reactor matrix is also a potential issue. In this report, we disclose an Advanced Moderator Module (AMM) concept enabled by an innovative enclosure solution combining the advantages of refractory metals, ceramic matrix composites (CMCs), and advanced coating technology to serve as hydrogen permeation barrier up to very-high temperatures. A schematic description of the AMM structure is illustrated in Figure A1. The AMM contains a moderating material core made of metal hydride with high thermal stability, such as YH 2-x . The hydride core is enclosed by an H 2 barrier layer coated on a ductile refractory metal liner to minimize hydrogen loss during high-temperature operation. A ceramic matrix composite (CMC) cladding is adopted to provide further structural strength, especially during power transients. Between the CMC cladding and metal liner, an extra diffusion barrier coating is inserted to suppress the chemical interaction at elevated temperatures. Hence, based on a series of innovative material solutions, the AMM is capable of containing the metal hydride core at elevated temperature (>900oC) inside coated and lined CMC envelop with negligible hydrogen loss throughout the microreactor lifetime. The benefit of the AMM technology was assessed based on a comprehensive multi-stage reactor-physics analysis. Advanced moderation based on hydride metals (such as YH2-x) enables reaching optimum moderation with higher fuel content than traditional VHTR technology, which is required to design compact micro-reactor cores targeting long-life operation. The AMM enclosure technology provides lower thermal neutron absorption rates than traditional solutions (based on stainless steel or Mo-based alloy such as TZM), which reduces the fissile enrichment requirements by 6-8% on a TRISO-fueled design based on the EMPIRE core. Finally, combining the hydride moderator with neutron transparent enclosure solutions provides significant potential to boost neutronics performance of micro-reactors in terms of increased core lifetime or reduced size and weight by 30-50% on a micro-reactor based on the Holos Quad technology. The progress and plans for on-going development and demonstration efforts are also discussed in this report. The current demonstration of AMM is focused on the hydrogen diffusion barrier demonstration under thermal cycling, while radiation tolerance demonstration is planned. The next step of the demonstration plan will be the assembly of miniature AMM for high-temperature testing, which can be used as a prototype for future pilot scale demonstration.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2021

Ionizing radiation effects in SONOS-based neuromorphic inference accelerators

Here, we evaluate the sensitivity of neuromorphic inference accelerators based on Silicon-Oxide-Nitride-Oxide-Silicon (SONOS) charge trap memory arrays to total ionizing dose (TID) effects. Data retention statistics were collected for 16 Mbit of 40 nm SONOS digital memory exposed to ionizing radiation from a Co-60 source, showing good retention of the bits up to the maximum dose of 500 krad(Si). Using this data, we formulate a rate-equation-based model for the TID response of trapped charge carriers in the ONO stack, and predict the effect of TID on intermediate device states between ‘program’ and ‘erase’. This model is then used to simulate arrays of low-power, analog SONOS devices that store 8-bit neural network weights and support in situ matrix-vector multiplication. We evaluate the accuracy of the irradiated SONOS-based inference accelerator on two image recognition tasks – CIFAR-10 and the challenging ImageNet dataset – using state-of-the-art convolutional neural networks, such as ResNet-50. We find that across the datasets and neural networks evaluated, the accelerator tolerates a maximum TID between 10 krad(Si) and 100 krad(Si), with deeper networks being more susceptible to accuracy losses due to TID.

43 PARTICLE ACCELERATORS↗
DOE OSTI2021

Extraction of gold, mercury and lead on CL resin in HCl

The uptake behavior of 197 mHg, 194 Au and 212 Pb in HCl on CL resin (Triskem) was studied. The extraction of mercury and gold is high (D w > 5,000) at all concentrations of HCl, with a maximum extraction of gold at 4 M HCl (D w ~ 53,000) and maximum extraction of mercury at 0.01 M HCl (D w ~ 44,000). Lead has moderate uptake on CL resin (D w ~ 100) from lower concentrations of HCl (0.001 to 0.05 M) and no uptake on the resin in concentrations ≥ 0.4 M. The uptake kinetics are similar for all three isotopes with high weight distribution ratios (>1,000 for 197 mHg, 194 Au and > 100 for 212 Pb) in ~ 30 s and equilibrium established at ~ 5 min. A column study was performed to demonstrate that 197m Hg, 194 Au and 212 Pb could be retained on CL resin in dilute HCl and 212 Pb eluted in 1 M HCl with no detectable 197m Hg or 194 Au. Elution of 197 mHg and 194 Au was performed with concentrated HNO 3 , but with significant losses to the column, particularly for 194 Au.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2023

Anisotropic magnetism and electronic structure of trigonal EuAl 2 Ge 2 single crystals

Understanding the interplay between magnetic and electronic degrees of freedom is of profound recent interest in different Eu-based magnetic topological materials. Here, we studied the magnetic and electronic properties of the layered Zintl-phase compound EuAl 2 Ge 2 crystallizing in the trigonal CaAl 2 Si 2 -type structure. We report zero-field neutron diffraction, temperature T - and magnetic-field H-dependent magnetic susceptibility χ(T, H), isothermal magnetization M(T, H), heat capacity C p (T, H), and electrical resistivity ρ(T, H) measurements, together with T-dependent angle-resolved photoemission spectroscopy (ARPES) measurements complemented with first-principle calculations. EuAl 2 Ge 2 undergoes second-order A-type antiferromagnetic (AFM) ordering below T N = 27.5(5) K, with the Eu moments (Eu 2+ , S = 7/2) aligned ferromagnetically in the ab plane while these layers are stacked antiferromagnetically along the c axis. The critical fields at which all moments become parallel to the field are 37.5(5) and 52.5(5) kOe for H ∥ ab and H ∥ c, respectively. The H = 0 magnetic structure consists of trigonal AFM domains associated with ab-plane magnetic anisotropy and a field-induced reorientation of the Eu spins in the domains is also evident at T = 2 K below the critical field H c1 = 2.5(1) kOe. The ρ(T) measurements reveal metallic behavior transforming into a slight resistivity increase on cooling towards T N . A pronounced loss of spin-disorder scattering is observed below T N . The ARPES results show that EuAl 2 Ge 2 is metallic both above and below T N , and the Fermi surface is anisotropic with two hole pockets at the zone center and one small electron pocket at each M point. In the AFM phase, we directly observe folded bands in ARPES due to the doubling of the magnetic unit cell along the c axis with an enhancement of quasiparticle weight due to the complex change in the coupling between the magnetic moments and itinerant electrons on cooling below T N . The observed electronic structure is well reproduced by first-principle calculations, which also predict the presence of nontrivial electronic states near the Fermi level in the AFM phase with Z 2 topological numbers 1;(000)

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2020

Non-catalytic oxidative depolymerization of lignin in perfluorodecalin to produce phenolic monomers

We demonstrate for the first time non-catalytic, oxidative cracking with molecular oxygen (O 2 ) to depolymerize native lignin into oxygenated phenolic monomers. Maximum monomer yield of 10.5 wt% was achieved at 250 °C after only 10 min of reaction and included vanillin, syringaldehyde, vanillic acid, and syringic acid. High rates of oxidation are attributed to the use of perfluorodecalin as solvent. Perfluorodecalin is a perfluorocarbon (PFC), characterized by their chemical stability and exceptionally high solubility for O 2 . Monomer yields were typically five-fold higher in perfluorodecalin compared to solvents more commonly employed in lignin conversion, such as methanol, butanol, acetonitrile, and ethyl acetate. Phenolic monomer production in perfluorodecalin favors high temperatures and short reaction times to prevent further oxidation of the produced monomers. Lignin oil obtained under oxidative conditions in perfluorodecalin showed lower molecular weight and smaller polydispersity compared to other solvents. Increasing the reaction time further decreased the molecular weight, while increasing reaction time in an inert atmosphere increased the molecular weight of the lignin oil. High concentrations of O 2 in perfluorodecalin not only increased lignin depolymerization but suppressed undesirable condensation reactions. Depolymerization is likely initiated by thermally induced homolytic cleavage of ether linkages in lignin to form phenoxy and carbon-based radicals. These radicals bind with O 2 as a radical scavenger and further react to form phenolic monomeric products rather than repolymerizing to large oligomers. The PFC process was scaled from 5 mL to 250 mL without any loss of yield. Because most organic compounds are not soluble in perfluorodecalin, recycling is easily achieved via liquid–liquid separation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Distributed quantum sensing enhanced by continuous-variable error correction

Abstract A distributed sensing protocol uses a network of local sensing nodes to estimate a global feature of the network, such as a weighted average of locally detectable parameters. In the noiseless case, continuous-variable (CV) multipartite entanglement shared by the nodes can improve the precision of parameter estimation relative to the precision attainable by a network without shared entanglement; for an entangled protocol, the root mean square estimation error scales like 1/ M with the number M of sensing nodes, the so-called Heisenberg scaling, while for protocols without entanglement, the error scales like 1 / M . However, in the presence of loss and other noise sources, although multipartite entanglement still has some advantages for sensing displacements and phases, the scaling of the precision with M is less favorable. In this paper, we show that using CV error correction codes can enhance the robustness of sensing protocols against imperfections and reinstate Heisenberg scaling up to moderate values of M . Furthermore, while previous distributed sensing protocols could measure only a single quadrature, we construct a protocol in which both quadratures can be sensed simultaneously. Our work demonstrates the value of CV error correction codes in realistic sensing scenarios.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

Achieving a high-performance sodium-ion pouch cell by regulating intergrowth structures in a layered oxide cathode with anionic redox

In P2-type layered transition metal (TM) oxides, which are typical cathode materials for Na-ion batteries, the presence of Li within the TM layer could lead to the formation of specific Na–O–Li configurations that trigger additional oxygen redox at high charging voltages. However, the prismatic-type (P-type) to octahedral-type (O-type) phase transition and irreversible TM migration could be simultaneously aggravated in high state of charge, resulting in structural distortion. Here we demonstrate that excessive desodiation of P2-Na 0.67 Li 0.1 Fe 0.37 Mn 0.53 O 2 (NLFMO) induces the formation of neighbouring O-type stacking faults with an intergrowth structure (that is, interlacing of O- and P-type layers), which leads to out-of-lattice Li migration and irreversible oxygen loss. We show that, by controlling the depth of charge to tailor the intergrowth structure, a P-type stacking state can be uniformly interspersed between the O-type stacking state, thereby avoiding neighbouring O-type stacking faults. Adjusting the O/P intergrowth structure leads to both reversible migration of Li/TM ions and reversible anionic redox in the NLFMO cathode. In conclusion, we thereby achieve a high-performance pouch cell (with an energy density of 165 W h kg −1 based on the entire weight of the cell) with both cationic and anionic redox activities.

25 ENERGY STORAGE↗
DOE OSTI2022

Network Anomaly Detection Using Federated Learning

The internet is turning out to be an integral part of every-one's lives as more and more devices are being connected to serve societal needs. Our work is motivated by two ma-jor observations. Firstly, one drawback of connecting to the network is the threat of network attacks that can compromise users' private information, leading to data loss and adversely affecting productivity. There are several traditional security mechanisms to defend against these attacks, such as firewalls, virtual private networks (VPNs), demilitarized zones (DMZs), and vulnerability scanners. One way to prevent these attacks is early detection and prevention. However, these kinds of architecture do not scale very well because of their centralized nature. Secondly, we observe from heuristics and data set distributions that the majority of the requests made to a server are innocuous. Therefore, almost all server request data sets are highly imbalanced, weighted highly towards the harmless requests.

Marfo, William↗
DOE OSTI2025

Tailored glycolysis of Nylon 6 to enable upcycling into high-strength adhesives

Nylon is a high-strength polyamide widely used in automotive, textiles, packaging, etc. However, its durability makes nylon waste difficult to manage, with recycling limited to mechanical grinding to make fillers. Here, we report a catalytic glycolysis approach to deconstruct Nylon 6 into controlled-length oligomers, enabling upcycling into value-added materials. A low-molecular-weight oligomer (M n = 1.8 kg/mol) was repolymerized with diepoxy-terminated poly(bisphenol A-co-epichlorohydrin) via mechanochemistry to create a high-performance adhesive. This copolymer achieves lap shear strengths over 22 MPa on steel and bonds steel to carbon fiber composites, nearly tripling the performance of commercial adhesives even at 90°C. Thermomechanical analyses show that the adhesive retains thermal stability similar to nylon 6 but melts at lower temperatures, allowing easier processing. The material can be reprocessed and reused without significant performance loss. This study demonstrates a strategy to convert nylon 6 waste into valuable materials, offering a sustainable path for difficult-to-recycle plastics.

Nylon waste↗
DOE OSTI2022

Design of a 3rd Order Lumped-Element Filtering Attenuator

This paper presents a novel filtering attenuator (filtenuator) concept, which allows for adjustable passband attenuation and filtering characteristics in a single design. The mathematical equations relating the desired passband loss to required resistance values are provided and validated with ideal circuit simulations. A third-order Chebyshev static filtenuator is designed at 1 GHz with multiple values of attenuation using lumped-element components as a proof of concept. Simulated and measured results for a 3 dB filtenuator show good agreement and several attenuation levels are shown to validate the concept. The proposed static filtenuator has advantages of ease of modification, compact size, and low cost, which are necessities in meeting the needs of reducing size, weight, power, and cost (SWaP-C) in future radio frequency transceivers.

Filter, Attenuator, Band-pass Filter (BPF), Lumped↗
DOE OSTI2023

Characterization and molecular simulation of lignin in Cyrene pretreatment of switchgrass

Biomass-derived solvents have been proposed as a novel pathway in biorefining for the realization of biofuels and bioproducts derived from lignocellulosic biomass. Cyrene derived from cellulose has recently been shown to have a high potential as a green organic solvent for pretreating poplar biomass. However, due to its high dynamic viscosity nature, high Cyrene concentration could cause negative effects on the sugar release of the pretreated biomass as well as driving up the operational cost of the lignin recovery. In this study, we combine experimental and computational approaches to examine the impact of Cyrene pretreatment with reduced Cyrene concentration under mild conditions on switchgrass lignin. Our experimental studies indicated correlation between pretreatment condition and recovery and structure modification of lignin. Switchgrass lignin extracted by Cyrene pretreatment possessed high preservation of β-O-4 ether inter-unit linkage, which could provide versatility in the integration of downstream lignin valorization into the modern biorefinery industries. Molecular modeling examining the solvation of switchgrass lignin polymer and the disaggregation of low-molecular weight lignin aggregates under pretreatment conditions indicated that a preferential interaction exists between Cyrene and lignin, which likely drives lignin release, and that the disruption of inter-lignin contacts can be modulated as a non-monotonic function of Cyrene : water ratio. Further, while Cyrene–lignin interactions permit the solubilization of lignin, simulations with proxy reactive-species reveal that changes to the diffusion of these reactive proxies and their localization near linkage sites under Cyrene conditions may inhibit chemical processes. In conclusion, the results indicated that loss of pretreatment efficacy caused by low Cyrene concentration could be compensated by prolonged pretreatment time and high catalyst dosage.

09 BIOMASS FUELS↗
DOE OSTI2026

Regularization via f -Divergence: An Application to Multi-Oxide Spectroscopic Analysis

In this paper, we explore the application of convolutional neural networks (CNNs) for predicting the chemical composition of complex geologic samples in a simulated Martian atmospheric environment. Specifically, we aim to characterize oxide weight percentages (wt.%) of rock samples analyzed by remote Laser-Induced Breakdown Spectroscopy (LIBS), framing the problem as a multi-target regression task . Neural networks trained on LIBS spectra are prone to overfitting due to high spectral complexity, limited labeled data, and measurement noise. While regularization is critical for improving generalization, common methods (e.g., ℓ 2 regularization) impose constraints not directly tied to data distribution properties. We propose a novel regularization method based on a specific ƒ-divergence induced by a graph-based estimator, designed to constrain the distributional discrepancy between predictions and targets. This regularizer serves a dual purpose: (a) mitigating overfitting by enforcing a constraint on the distributional difference between predictions and noisy targets, and (b) acting as an auxiliary loss that penalizes large divergences. To enable backpropagation, we develop a differentiable approximation of this particular ƒ-divergence, making the method feasible for neural networks. Experiments on ChemCam and SuperCam LIBS calibration spectra show that mathematical equation-divergence regularization outperforms or matches standard regularization methods (ℓ 1 , ℓ 2 , dropout) and the classical baseline, partial least squares (PLS). Combining ƒ-divergence regularization with standard regularization yields further performance gains, indicating that distributional regularization is useful in this context giving a promising direction for robust model training in planetary science applications. Source code is publicly available at Klein and Li (2025), https://doi.org/10.11578/dc.20250530.7.

58 GEOSCIENCES↗