SEARCH · Engineering Papers

Results for “Batch processing”

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 757 records · Page 42

All sources NASA NTRS DOE OSTI

DOE OSTI2017

Contradictory effects of silver nanoparticles on activated sludge wastewater treatment

Increased amount of nano-silver will be released into domestic and industrial waste streams due to its extensive application. However, great controversy still exists on the effects of silver nanoparticle (Ag-NP) on biological wastewater treatment processes and a toxicology model has not been built yet. Four sequencing batch reactors with activated sludge has been run for over three months with different silver species at a concentration of 1 mg Ag/L in influent. Both freshly prepared Ag-NPs and aged Ag-NPs were tested with released silver ion as control. Results showed that Ag-NPs, especially freshly prepared Ag-NPs, can help to maintain or even increase the diversity of microbial community in activated sludge and the biomass concentration even under long-term treatment. It indicates that the hormesis model need to be considered for the toxicology of Ag-NPs.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2021

Air oxidation of yttrium hydride as a high temperature moderator for thermal neutron spectrum fission reactors

Yttrium hydride (YH x ) is an attractive moderator material for thermal neutron spectrum fission reactors requiring a small reactor core volume and has been selected as the neutron moderator for the Transformational Challenge Reactor (TCR), an advanced gas-cooled microreactor. Before YH x can be used in this application, it is important to understand the material response to off-normal conditions. In the present study, 550–650 °C isothermal dry air oxidation was performed to simulate a depressurized loss of force circulation (DLOFC) event. The oxidation was performed using thermogravimetric analysis (TGA) on bulk crack-free YHx coupons. Oxidation studies were also performed on Y coupons to elucidate the impact of H on oxidation. Both the chemistry and distribution of processing impurities were found to strongly affect oxidization behavior on a batch-to-batch basis. Regardless of batch, YHx oxidized at a significantly lower rate than Y at all temperatures, and the lower rate was directly correlated with increased hydride content. Metallic Y exhibited complex exponential kinetics, whereas YH x also exhibited complex kinetics but gained considerably less mass. According to literature reports on protonic and native-ion conductivities of Y 2 O 3 and mass spectrometry analysis of gaseous reaction products formed during the oxidation of YH x , a mechanism for the reduced oxidation rate of yttrium hydride is suggested.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2015

Kinetics of hydrogen isotope exchange in β-phase Pd-H-D

Hydrogen isotope gas exchange within palladium powders is examined using a batch-type reactor coupled to a residual gas analyzer (RGA). Furthermore, the exchange rates in both directions (H 2 + PdD and D 2 + PdH) are measured in the temperature range 178–323 K for the samples with different particle sizes. The results show this batch-type exchange is closely approximated as a first-order kinetic process with a rate directly proportional to the surface area of the powder particles. An exchange rate constant of 1.40 ± 0.24 μmol H 2 /atm cm 2 s is found for H 2 + PdD at 298 K, 1.4 times higher than that for D 2 + PdH, with an activation energy of 25.0 ± 3.2 kJ/mol H for both exchange directions. Finally, a comparison of exchange measurement techniques shows these coefficients, and the fundamental exchange probabilities are in good agreement with those obtained by NMR and flow techniques.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2018

Post-Fermentation Recovery of Biobased Carboxylic Acids

Carboxylic acids are common products produced from the bioconversion of renewable feedstocks. In these processes the separation of the acid product from fermentation broth is the most energy and cost intensive unit operation. Thus, the development of robust, scalable separation approaches that can be applied to a variety of carboxylates is of critical importance to the development of processes that utilize carboxylic acids as platform chemicals. Here we report a batch separation method that includes cell and particulate removal, cation exchange, activated carbon treatment, dewatering with a polymer resin, and product recovery. This method is demonstrated on two unique fermentation broths both derived from corn stover hydrolysate to separate neat succinic and propionic acid. For succinic acid, a crystallization yield of 91% with a product purity of 99.93% was achieved. To our knowledge this is the highest reported crystallization yield and purity for the recovery of succinic acid. Additionally, the method requires approximately 50% less energy compared to standard evaporative crystallization approaches. For propionic acid, neat liquid product was obtained with a distillation yield of 80% and purity of 98%. Finally, these excellent results achieved in terms of yield and purity for succinic and propionic acid, two acids with widely different physical properties, from chemically complex hydrolysate broth demonstrates the effective and robust nature of this approach.

09 BIOMASS FUELS↗

DOE OSTI2020

Impacts of the Addition of Sodium Reactor Experiment (SRE) and DR-3 Fuel from H-Canyon to Tank 40 on Acid Stoichiometry and Hydrogen Production

H-Canyon plans to discard Sodium Reactor Experiment (SRE) material that is currently stored in Tanks 16.3 and 16.4. Savannah River Remediation (SRR) determined that a portion of the SRE material can be added to Sludge Batch 9 (SB9) in Tank 40 and the remainder will be added to Sludge Batch 10 (SB10) that is currently being prepared in Tank 51. A processing flowsheet for both the nitric-formic acid flowsheet and the nitric-glycolic acid flowsheet were previously developed for SB9 sludge-only and coupled operation with the Actinide Removal Process – Modular Caustic Side Solvent Extraction Unit (ARP-MCU) and the Salt Waste Processing Facility (SWPF). Due to the compositional change in Tank 40 after the addition of SRE material, it is necessary to determine the influence on the acid stoichiometry operating window and hydrogen production for SB9 processing. An assessment was completed to address the influence of an addition of SRE and DR-3 Fuel from H-Canyon on the chemistry of Tank 40, the defined acid stoichiometry window (with and without SWPF streams) for the nitric-formic and nitric-glycolic acid flowsheets, and the resulting influence of the hydrogen production. Hydrogen production during the SRAT/SME cycles consists of thermolytic, radiolytic, and catalytic hydrogen. For the nitric-formic acid flowsheet, catalytic hydrogen dominates the hydrogen production during DWPF operation mode for the SRAT and SME vessels. This report was written to summarize that assessment. No SRNL testing is needed prior to adding up to 7,100 additional gallons of SRE and DR-3 Fuel from H-Canyon to Tank 40. Also, the addition of SRE will not exceed the Shielded Cells reported values for catalytic hydrogen for either the nitric-formic acid flowsheet (Technical Safety Requirement (TSR) limit of 0.15 lb/h) or the nitric-glycolic acid flowsheet (TSR limit 0.024 lb/h). The addition of the ARP stream from SWPF and the use of the nitric-glycolic acid flowsheet are both expected to decrease hydrogen generation, leading to a larger safety margin in the CPC.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2023

Investigation of potential causes of production limiting fouling within SWPF due to mercury solids formation during pH swings within the SWPF process

The Salt Waste Processing Facility (SWPF) goal to investigate the use a high sodium Salt Batch 5 Qualification sample as the radioactive salt solution for mixing tests meant to evaluate the propensity for solids formation between the 5.6 M sodium clarified salt solution (CSS) and scrub solutions (0.05 M HNO 3 and 0.025 M NaOH solutions) during pH changes has been completed at SRNL. The key results and conclusions from this investigation to evaluate the propensity for solids formation between CSS and scrub solutions because of pH fluctuations include the following: None of the 8 test matrices, involving the mixing/titrating of scrub solution with CSS and vice versa, showed any visible evidence for the formation of solids with pH changes. The average turbidity measurements for these 8 tests, excluding test 7, was 21.6 ± 0.4 NTU. This average turbidity value is within 2 sigma of the baseline turbidity value of 20.5 NTU for the CSS only without added scrub solutions. However, only in test 7, which involved the titrating of CSS into dilute nitric acid scrub solution (0.05 M HNO 3 solution), was a small measurable change in the turbidity of the mixture observed at 31.9 ºC. On the other hand, the equivalent test (instantaneous addition of CSS into 0.05 M HNO 3 ) gave an average turbidity measurement value of 21.6 ± 0.6 NTU at a temperature of 24.0 ºC. This difference in turbidity measurement has been attributed to either instrument performance errors or the effect of temperature. It is recommended that these CSS and scrub solution mixing tests be performed with corresponding simulant solutions in a non-radioactive environment. A simulant environment will aid in the full integration of a pH and turbidity meter equipment with the experimental set up. A new detection method, preferable a laser-based method, may be needed in detecting and confirming the precipitation of fine particulates at the right blending times and pH fluctuations in the solution mixtures. On the other hand, the differing free OH concentrations (~2 M in CSS), H + concentrations (0.05 M HNO 3 ) for the titrants and the bulk solutions, may make it almost impossible to observe a neutral pH or a pH in the range of 7-9, where the potential for solids formation is maximized and leads to possible precipitation of minerals like aluminum hydroxides.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2020

COMBLE-ISLAS water isotopes in precipitation (COMBLEISLASISO)

During the measurement period, precipitation was sampled on daily to sub-daily basis with a sampling kit consiting of a clear plastic box (60x40x40 cm) mounted to a ground structure. Daily sampling lasted from 22 Feb to 24 Mar 2020. At a fixed time of day (10 UTC), samples were collected and the sampling box cleaned for the next sample. Additionally, during intense observations periods (IOPs) announced by ISLAS, ARMS scientists increased sampling frequency to sub-daily sampling. During the campaign period, 2 IOPs were conducted, with in total 12 samples taken. IOP1 lasted from 27 Feb 10 to 02 Mar 2020, and IOP2 lasted from 12 Mar to 14 Mar 2020. Solid precipitation in the box was then melted, transferred to an 8 ml glass vial, sealed and shipped to FARLAB (University of Bergen, Norway) for stable water isotope analysis. Measured stable isotope composition of precipitation samples will be related to conditions at other sampling locations, and at the evaporation site to inform about conservation of water isotope quantities. In the period from 22 Feb to 24 Mar, 23 daily samples were collected at 10 UTC for stable water isotope analysis. During the entire sampling period, a high variability of precipitation phase was encountered. Most samples were recovered in the form of graupel/snow (12 samples), as rain (7 samples), and the remaining samples fell as graupel, snow, rain or frozen rain. Precipitation amounts varied between <1 mm (26 Feb) and 140 mm (25 Feb) within the sampling box. Furthermore, higher-resolution sampling was conducted during 2 intense observation periods (IOPs). During IOP1, lasting from 27 Feb 10 UTC to 02 Mar 10 UTC, 7 samples were collected, all of them as solid phase precipitation. IOP2 lasted from 12 Mar 10 UTC to 14 Mar 09 UTC, and 5 samples (all as graupel/snow) were collected. Samples were processed according to FARLAB standard measurement procedures. In short, samples were transferred to 1.5 ml glass vials with rubber/PTFE septa (part #548-0907, VWR, USA). An autosampler (A0325, Picarro Inc) transferred ca. 2µl per injection into a high-precision vapourizer (A0211, Picarro Inc, USA) heated to 110°C. After blending with dry N2 (< 5 ppm H2O) the gas mixture was directed into the measurement cavity of a Cavity-Ring Down Spectrometer (L2140-i, Picarro Inc) for about 7 min with a typical water concentration of 20 000 ppm. Memory effects were reduced by two times measuring a vapour mixture at a mixing ratio of 50 000 ppm, obtained from 2 injections of 2 µl for 5 min at the beginning of each new sample vial. Thereafter, another 6 injections of 2 µl per sample were measured individually as described above, and averages of the last 5 injections were used for further processing. Three standards were measured at the beginning and end of each batch, including a drift standard DI2 (δD: -50.72±0.73 permil, δ18O: -7.63±0.10 permil), and for calibration the laboratory standards GLW (δD: -307.79±0.75 permil, δ18O: -40.02±0.07 permil) and EVAP2 (δD: 9.52±0.65 permil, δ18O: 1.81±0.13 permil). A detailed calibration report is included with the final uploaded data set.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1981

Ultrapure glass optical waveguide development in microgravity by the sol-gel process

Multicomponent, homogeneous, noncrystalline oxide gels can be prepared by the sol-gel process and these gels are promising starting materials for melting glasses in the space environment. The sol-gel process referred to here is based on the polymerization reaction of alkoxysilane with other metal alkoxy compounds or suitable metal salts. Many of the alkoxysilanes or other metal alkoxides are liquids and thus can be purified by distillation. The use of gels offers several advantages such as high purity and lower melting times and temperatures. The sol-gel process is studied for utilization in the preparation of multicomponent ultrapure glass batches for subsequent containerless melting of the batches in space to prepare glass blanks for optical waveguides.

Mukherjee, S. P.↗

NASA NTRS1981

Development and demonstration of manufacturing processes for fabricating graphite/Larc-160 polyimide structural elements, part 4, paragraph B

Progress in the development of processes for production of Celion/LARC-160 graphite-polyimide materials, quality control, and the fabrication of Space Shuttle composite structure components is reported. Liquid chromatographic analyses of three repeatibility batches were performed and are compared to previous Hexcel standard production and to variables study LARC-160 intermediate resins. Development of processes for chopped fiber molding are described and flexural strength, elastic modulus, and other physical and mechanical properties of the molding are presented.

Source record↗

NASA NTRS2008

NASA Tech Briefs, February 2008

Topics discussed include: Optical Measurement of Mass Flow of a Two-Phase Fluid; Selectable-Tip Corrosion-Testing Electrochemical Cell; Piezoelectric Bolt Breakers and Bolt Fatigue Testers; Improved Measurement of B(sub 22) of Macromolecules in a Flow Cell; Measurements by a Vector Network Analyzer at 325 to 508 GHz; Using Light to Treat Mucositis and Help Wounds Heal; Increasing Discharge Capacities of Li-(CF)(sub n) Cells; Dot-in-Well Quantum-Dot Infrared Photodetectors; Integrated Microbatteries for Implantable Medical Devices; Oxidation Behavior of Carbon Fiber-Reinforced Composites; GIDEP Batching Tool; Generic Spacecraft Model for Real-Time Simulation; Parallel-Processing Software for Creating Mosaic Images; Software for Verifying Image-Correlation Tie Points; Flexcam Image Capture Viewing and Spot Tracking; Low-Pt-Content Anode Catalyst for Direct Methanol Fuel Cells; Graphite/Cyanate Ester Face Sheets for Adaptive Optics; Atomized BaF2-CaF7 for Better-Flowing Plasma-Spray Feedstock; Nanophase Nickel-Zirconium Alloys for Fuel Cells; Vacuum Packaging of MEMS With Multiple Internal Seal Rings; Compact Two-Dimensional Spectrometer Optics; and Fault-Tolerant Coding for State Machines.

Source record↗

DOE OSTI2019

Laboratory-Scale Melter Systems for Simulated and Real AP-105 Tank Waste - 19301

The Hanford Tank Waste Treatment and Immobilization Plant is being constructed to vitrify the nuclear waste stored in underground tanks on the Hanford Site. Through this conversion process, waste components become partitioned into one of two product streams: glass and off-gas, including condensate. A small-scale melter system is appropriate for analyzing the waste-to-glass process because it does not require the high volumes of input waste or resource commitment of a large-scale melter system while providing dynamic information that can be difficult to determine from batch reactions in a crucible system. A continuous laboratory-scale melter (CLSM) was sized to collect the relevant process and chemical information from obtainable volumes of Hanford tank waste and designed to operate with a continuous feeding process while periodically pouring glass product and collecting off-gas. Two CLSMs were constructed: one in a non-radioactive environment for processing waste simulants and another in a fume hood capable of handling radioactive material for processing actual Hanford tank waste. Multiple waste simulants were first tested in the non-radioactive CLSM to determine the processability and desired operating window for actual tank waste. About 12 liters of real waste from Hanford tank 241-AP-105 were received, filtered, and ion exchanged for removal of Cs-137. This quantity of waste was vitrified into 9.5 kg of glass product over the course of ∼15 hours of feeding to the CLSM. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

NASA NTRS1988

Improved silicon carbide for advanced heat engines

This is the third annual technical report for the program entitled, Improved Silicon Carbide for Advanced Heat Engines, for the period February 16, 1987 to February 15, 1988. The objective of the original program was the development of high strength, high reliability silicon carbide parts with complex shapes suitable for use in advanced heat engines. Injection molding is the forming method selected for the program because it is capable of forming complex parts adaptable for mass production on an economically sound basis. The goals of the revised program are to reach a Weibull characteristic strength of 550 MPa (80 ksi) and a Weibull modulus of 16 for bars tested in 4-point loading. Two tasks are discussed: Task 1 which involves materials and process improvements, and Task 2 which is a MOR bar matrix to improve strength and reliability. Many statistically designed experiments were completed under task 1 which improved the composition of the batches, the mixing of the powders, the sinter and anneal cycles. The best results were obtained by an attritor mixing process which yielded strengths in excess of 550 MPa (80 ksi) and an individual Weibull modulus of 16.8 for a 9-sample group. Strengths measured at 1200 and 1400 C were equal to the room temperature strength. Annealing of machined test bars significantly improved the strength. Molding yields were measured and flaw distributions were observed to follow a Poisson process. The second iteration of the Task 2 matrix experiment is described.

Whalen, Thomas J.↗

DOE OSTI2022

Scaling up carboxylic acid production from cheese whey and brewery wastewater via methane-arrested anaerobic digestion

In a circular economy, organic waste streams are valuable resources for sustainably producing chemicals and fuels. This study investigates a new methane-arrested anaerobic digestion (MAAD) process that converts high-strength cheese whey and brewery wastewater into carboxylic acids. The process was developed and optimized under various bench-scale semi-continuous (fed-batch) operating conditions (e.g., retention time, organic loading rate, pH, and feed/harvest frequency). The MAAD responses to various control-failure scenarios were also systematically investigated. The highest total acid productivity was 26g/(L liq ·d) with a substrate conversion of 0.79g COD digested /g COD fed at a hydraulic retention time (HRT) of approximately 2 d in a 14-L digester. The most stable conditions for digester operation (HRT 3 d at pH 6.0 and 40°C) were selected for process scale-up to 100gal (~380 L). Semi-continuous, pilot-scale MAAD successfully produced a total acid concentration of 40.6+/-1.1 g/L with 8.1% acetic acid, 45.1% butyric acid, and 44.5% lactic acid. The links between wastewater characteristics, operation mode, digester scale, and microbial community structure were statistically analyzed. The results show genera Sporolactobacillus and Clostridium positively correlate with butyric acid production (Pearson correlation coefficient>0.5). Moreover, four kinetic models were developed and fit to batch MAAD experimental datasets (R 2 >95%) and were successfully applied to predict the total acid production in both bench-and pilot-scale semi-continuous MAADs. In conclusion, this study shows MAAD has the potential for industrial-scale applications and is a robust platform to valorize low- or negative-value waste streams into high-value bioproducts.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2023

Accelerating Machine Learning Inference with GPUs in ProtoDUNE Data Processing

Abstract We study the performance of a cloud-based GPU-accelerated inference server to speed up event reconstruction in neutrino data batch jobs. Using detector data from the ProtoDUNE experiment and employing the standard DUNE grid job submission tools, we attempt to reprocess the data by running several thousand concurrent grid jobs, a rate we expect to be typical of current and future neutrino physics experiments. We process most of the dataset with the GPU version of our processing algorithm and the remainder with the CPU version for timing comparisons. We find that a 100-GPU cloud-based server is able to easily meet the processing demand, and that using the GPU version of the event processing algorithm is two times faster than processing these data with the CPU version when comparing to the newest CPUs in our sample. The amount of data transferred to the inference server during the GPU runs can overwhelm even the highest-bandwidth network switches, however, unless care is taken to observe network facility limits or otherwise distribute the jobs to multiple sites. We discuss the lessons learned from this processing campaign and several avenues for future improvements.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

NASA NTRS1995

Rapid Airplane Parametric Input Design (RAPID)

RAPID is a methodology and software system to define a class of airplane configurations and directly evaluate surface grids, volume grids, and grid sensitivity on and about the configurations. A distinguishing characteristic which separates RAPID from other airplane surface modellers is that the output grids and grid sensitivity are directly applicable in CFD analysis. A small set of design parameters and grid control parameters govern the process which is incorporated into interactive software for 'real time' visual analysis and into batch software for the application of optimization technology. The computed surface grids and volume grids are suitable for a wide range of Computational Fluid Dynamics (CFD) simulation. The general airplane configuration has wing, fuselage, horizontal tail, and vertical tail components. The double-delta wing and tail components are manifested by solving a fourth order partial differential equation (PDE) subject to Dirichlet and Neumann boundary conditions. The design parameters are incorporated into the boundary conditions and therefore govern the shapes of the surfaces. The PDE solution yields a smooth transition between boundaries. Surface grids suitable for CFD calculation are created by establishing an H-type topology about the configuration and incorporating grid spacing functions in the PDE equation for the lifting components and the fuselage definition equations. User specified grid parameters govern the location and degree of grid concentration. A two-block volume grid about a configuration is calculated using the Control Point Form (CPF) technique. The interactive software, which runs on Silicon Graphics IRIS workstations, allows design parameters to be continuously varied and the resulting surface grid to be observed in real time. The batch software computes both the surface and volume grids and also computes the sensitivity of the output grid with respect to the input design parameters by applying the precompiler tool ADIFOR to the grid generation program. The output of ADIFOR is a new source code containing the old code plus expressions for derivatives of specified dependent variables (grid coordinates) with respect to specified independent variables (design parameters). The RAPID methodology and software provide a means of rapidly defining numerical prototypes, grids, and grid sensitivity of a class of airplane configurations. This technology and software is highly useful for CFD research for preliminary design and optimization processes.

Smith, Robert E.↗

NASA NTRSDate not supplied

Optimizing Single Nuclei Sequencing of Brain Samples From Space Flown Mice Across Age and Strain

The NASA GeneLab Sample Processing Laboratory offers high-throughput sequencing services to NASA-funded space biology researchers. Space biology studies have specific challenges such as low sample numbers, introducing susceptibility to batch effects from sample handling. These issues are compounded by complex protocols such as single-nuclei isolation and sequencing, which has recently become an attractive methodology for assessing the cellular diversity within spaceflight samples. High quality single-nuclei sequencing requires reproducible protocols to dissociate tissue and generate clean suspension of intact single nuclei. Producing single-nuclei suspension from brain tissue is particularly challenging due to cell type heterogeneity and the myelin sheath that carries over into the nuclei suspension as debris. Current procedures tend to be time consuming and sometimes include steps that can alter gene expression and create cell-type bias. Commercially available nuclei isolation kits, such as the 10X Genomics nuclei isolation kit, offers a streamlined way to process samples for nuclei isolation, thereby minimizing batch effects and enabling reproducibility. In this study, we report on the performance of the 10X Genomics nuclei isolation kit and Chromium Next GEM Single Cell Multiome ATAC + Gene Expression kit to generate sequencing libraries from space-flown mouse brain samples. Single nuclei sequencing was performed on frozen mouse brain tissue from two spaceflight missions, Rodent Research-10 (RR-10) and RR Reference Mission-2 (RRRM-2). RR-10 mice were female B6129SF2/J, euthanized at 18-19 weeks whereas RRRM-2 mice were female C57BL/6NTac, euthanized at 20 or 37 weeks. Sequencing data was processed using standard GeneLab data processing pipelines. We report evaluation of the performance of the 10X Genomics nuclei isolation kit for spaceflight samples from mouse brain, and evaluation of reproducibility across different mouse strains and age groups. We also report preliminary scientific results including cell type inference, cell clustering, and differentially expressed genes and pathways between spaceflight and ground control samples.

RR-10↗

DOE OSTI2026

Active Learning‐Driven Inkless Additive Nanomanufacturing for Printed Electronics

Inkless additive nanomanufacturing for printed electronics promises broad material and substrate versatility, yet the high-dimensional print parameter space makes tuning print parameters time-intensive. We present a Bayesian optimization study that constructs a digital twin from printed-silver data to benchmark surrogate models, acquisition functions, and batch sizes head-to-head to achieve user-specified target resistance. Tested surrogate models included Gaussian process, random forest, and Bayesian neural network surrogates with expected improvement and confidence bound acquisition functions. In total, we evaluate 48 unique model configurations alongside a random sampling baseline for comparison. For printed silver, the Bayesian neural network with a batch size of one achieved the lowest average cumulative regret, approximately four times more efficient on average than random sampling. To balance performance and substrate space, a random forest model with expected improvement and a batch size of four was chosen as the model for validation testing. Applying this chosen configuration to copper with an additional print parameter, the model achieved a resistance within 0.15 Ω of a 1 Ω target in fewer than 30 printed lines across five validation sets. Altogether, the workflow yields a tuned and validated model that efficiently guides experiments toward the target while simultaneously learning the parameter space.

Bevel, Colton [Auburn University, AL (United State↗

DOE OSTI2025

Artificial intelligence driven laser parameter search: Inverse design of photonic surfaces using greedy surrogate-based optimization

Photonic surfaces designed with specific optical characteristics are becoming increasingly crucial for novel energy harvesting and storage systems. The design of these surfaces can be achieved by texturing materials using lasers. The optimal adjustment of laser fabrication parameters to achieve target surface optical properties is an open challenge. Thus, we develop a surrogate-based optimization approach. Our framework employs the Random Forest algorithm to model the forward relationship between the laser fabrication parameters and the resulting optical characteristics. During the optimization process, we use a greedy, prediction-based exploration strategy that iteratively selects batches of laser parameters to be used in experimentation by minimizing the predicted discrepancy between the surrogate model’s outputs and the user-defined target optical characteristics. This strategy allows for efficient identification of optimal fabrication parameters without the need to model the error landscape directly. We demonstrate the efficiency and effectiveness of our approach on two synthetic benchmarks and two specific experimental applications of photonic surface inverse design targets. By calculating the average performance of our algorithm compared to other state of the art optimization methods, we show that our algorithm performs, on average, twice as well across all benchmarks. Additionally, a warm starting inverse design technique for changed target optical characteristics enhances the performance of the introduced approach.

97 MATHEMATICS AND COMPUTING↗