Search NASA⌕ Search

SEARCH · Search NASA

Results for “batched factorization”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2016

Implementation and Tuning of Batched Cholesky Factorization and Solve for NVIDIA GPUs

Many problems in engineering and scientific computing require the solution of a large number of small systems of linear equations. Due to their high processing power, Graphics Processing Units became an attractive target for this class of problems, and routines based on the LU and the QR factorization have been provided by NVIDIA in the cuBLAS library. This work addresses the situation where the systems of equations are symmetric positive definite. The paper describes the implementation and tuning of the kernels for the Cholesky factorization and the forward and backward substitution. Targeted workloads involve the solution of thousands of linear systems of the same size, where the focus is on matrix dimensions from 5 by 5 to 100 by 100. Due to the lack of a cuBLAS Cholesky factorization, execution rates of cuBLAS LU and cuBLAS QR are used for comparison against the proposed Cholesky factorization in this work. Execution rates of forward and backward substitution routines are compared to equivalent cuBLAS routines. Comparisons against optimized multicore implementations are also presented. Superior performance is reached in all cases.

Computer Science↗
DOE OSTI2015

Batched matrix computations on hardware accelerators based on GPUs

Scientific applications require solvers that work on many small size problems that are independent from each other. At the same time, the high-end hardware evolves rapidly and becomes ever more throughput-oriented and thus there is an increasing need for an effective approach to develop energy-efficient, high-performance codes for these small matrix problems that we call batched factorizations. The many applications that need this functionality could especially benefit from the use of GPUs, which currently are four to five times more energy efficient than multicore CPUs on important scientific workloads. This study, consequently, describes the development of the most common, one-sided factorizations, Cholesky, LU, and QR, for a set of small dense matrices. The algorithms we present together with their implementations are, by design, inherently parallel. In particular, our approach is based on representing the process as a sequence of batched BLAS routines that are executed entirely on a GPU. Importantly, this is unlike the LAPACK and the hybrid MAGMA factorization algorithms that work under drastically different assumptions of hardware design and efficiency of execution of the various computational kernels involved in the implementation. Thus, our approach is more efficient than what works for a combination of multicore CPUs and GPUs for the problems sizes of interest of the application use cases. The paradigm where upon a single chip (a GPU or a CPU) factorizes a single problem at a time is not at all efficient in our applications’ context. We illustrate all of these claims through a detailed performance analysis. With the help of profiling and tracing tools, we guide our development of batched factorizations to achieve up to two-fold speedup and three-fold better energy efficiency as compared against our highly optimized batched CPU implementations based on MKL library. Finally, the tested system featured two sockets of Intel Sandy Bridge CPUs and we compared with a batched LU factorizations featured in the CUBLAS library for GPUs, we achieve as high as 2.5× speedup on the NVIDIA K40 GPU.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

SchedInspector: A Batch Job Scheduling Inspector Using Reinforcement Learning

Improving the performance of job executions is an important goal of HPC batch job schedulers, such as minimizing job waiting time, slowdown, or completion time. Such a goal is often accomplished using carefully designed heuristics based on job features, such as job size and job duration. However, these heuristics overlook important runtime factors (e.g., cluster availability and waiting job patterns), which may vary across time and make a previously sound scheduling decision not hold any longer. In this study, we propose a new approach to incorporate runtime factors into batch job scheduling for better job execution performance. The key idea is to add a scheduling inspector on top of the base job scheduler to scrutinize its scheduling decisions. The inspector will take the runtime factors into consideration and accordingly determine the fitness of the scheduled job. It then either accepts the scheduled job or rejects it and asks the base schedulers to try again later. We realize such an inspector, namely SchedInspector, by leveraging the intelligence of reinforcement learning. Through extensive experiments, we show SchedInspector can intelligently integrate the runtime factors into various batch job scheduling policies, including the state-of-the-art one, to gain better job execution performance, such as smaller average bounded job slowdown (up to 69% better) or average job waiting time (up to 52% better), across various real-world workloads. We also show that although rejecting scheduling decisions may leave the resources idle hence affect the system utilization, SchedInspector is able to achieve the job execution performance improvement with marginal impact on the system utilization (typically less than 1%). We consider one key advantage of SchedInspector is it automatically learns to work with and improve existing job scheduling policies without changing them, which makes it promising to serve as a generic enhancer for various batch job scheduling policies.

Zhang, Di↗
DOE OSTI2020

Cesium Removal Performance Comparisons of Crystalline Silicotitanate Media Batches with Savannah River Site Waste Simulant

The Tank Closure Cesium Removal (TCCR) system uses ion exchange columns filled with Crystalline Silicotitanate (CST) media to process radioactive waste solutions for the removal of Cs137. TCCR currently focuses on dissolving Savannah River Site (SRS) Tank 10H waste (primarily sodium salt cake solids) within the tank followed by at-tank ion exchange column treatment. Plans are underway to prepare and install a second TCCR unit at SRS. Capacity and particle size differences exist between archived (IE-911) and more recently-prepared CST media batches (9120- B and 9140-B). Side-by-side comparison testing was performed to evaluate the cesium removal performance of each batch to aid in selecting the preferred CST batch and media characteristics to load into the second TCCR unit. Batch contact equilibrium and flow-through column tests have been conducted with three CST batches using an SRS Average Simulant. Simulant batch contact equilibrium cesium loading results for the three CST batches (including two lots of one batch) are provided in Table ES-1. The 35 °C data indicates that the archived IE-911 batch has a higher cesium capacity than recently-prepared CST media and that the minor TCCR CST 9120-B media lot (2099000035) has similar cesium removal performance to the major TCCR lot (2099000034). Tests conducted at 25 ºC for the archived IE-911 CST batch indicated lower cesium removal performance with this simulant batch than was observed recently with a different SRS Average Simulant batch. A dilution factor (DF) is typically utilized when modeling engineered CST media cesium loading performance to account for mass contributions from the binder material. In cases where CST performance is lower than expected, this factor includes corrections for the binder and for low performance. The ZAM (Zheng, Anthony, and Miller) Isotherm Model DF values are provided in Table ES-1 for each simulant batch contact result. DF values near 0.5 were determined for the 9120-B and 9140-B CST batch contact tests while DF values near 0.6 (20% higher) were determined for the tests with IE-911. These dilution factors are lower than recently observed with a different SRS Average Simulant batch (9140-B DF = 0.68; IE-911 DF = 1.0). Based on these results, it appears that some component in the simulant solution used for equilibrium testing may have resulted in reduced cesium loading on the CST media.

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

Closing the Nuclear Fuel Cycle with a Simplified Minor Actinide Lanthanide Separation Process (ALSEP) and Additive Manufacturing

Expanded low-carbon baseload power production through the use of nuclear fission can be enabled by recycling long-lived actinide isotopes within the nuclear fuel cycle. This approach provides the benefits of (a) more completely utilizing the energy potential of mined uranium, (b) reducing the footprint of nuclear geological repositories, and (c) reducing the time required for the radiotoxicity of the disposed waste to decrease to the level of uranium ore from one hundred thousand years to a few hundred years. A key step in achieving this goal is the separation of long-lived isotopes of americium (Am) and curium (Cm) for recycle into fast reactors. To achieve this goal, a novel process was successfully demonstrated on a laboratory scale using a bank of 1.25-cm centrifugal contactors, fabricated by additive manufacturing, and a simulant containing the major fission product elements. Americium and Cm were separated from the lanthanides with over 99.9% completion. The sum of the impurities of the Am/Cm product stream using the simulated raffinate was found to be 3.2 × 10 –3 g/L. The process performance was validated using a genuine high burnup used nuclear fuel raffinate in a batch regime. Separation factors of nearly 100 for 154 Eu over 241 Am were achieved. All these results indicate the process scalability to an engineering scale.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2016

Microcrystalline diamond cylindrical resonators with quality-factor up to 0.5 million

We demonstrate high quality-factor 1.5 mm diameter batch-fabricated microcrystalline diamond cylindrical resonators (CR) with quality-factors limited by thermoelastic damping (TED) and surface loss. Resonators were fabricated 2.6 and 5.3 μm thick in-situ boron-doped microcrystalline diamond films deposited using hot filament chemical vapor deposition. The quality-factor (Q) of as-fabricated CR's was found to increase with the resonator diameter and diamond thickness. Annealing the CRs at 700 °C in a nitrogen atmosphere led to a three-fold increase in Q, a result we attribute to thinning of the diamond layer via reaction with residual O{sub 2} in the annealing furnace. Post-anneal Q exceeding 0.5 million (528 000) was measured at the 19 kHz elliptical wineglass modes, producing a ring-down time of 8.9 s. A model for Q versus diamond thickness and resonance frequency is developed including the effects of TED and surface loss. Measured quality factors are shown to agree with the predictions of this model.

36 MATERIALS SCIENCE↗
DOE OSTI2018

Analysis of the IBL and LBNL irradiated PIN and PN diodes

This report is a follow-up to the previous report on the difference between high fluence, high and low flux irradiations. There was a discrepancy in the data for the LBNL irradiated S5821 PIN diodes. There were diodes irradiated in the two batches (high and low flux) with the same flux and fluence for reference (lell ions/cm 2 /shot and 5, 10, and 20 ions/cm 2 total flux). Although these diodes should have the same electrical characteristics their leakage currents were different by a factor of 5-6 (batch 2 was larger). Also, the C-V measurements showed drastically different results. It was speculated that these discrepancies were due to one of the following two reasons: 1. Different times elapsed between radiation and characterization. 2. Different areas were irradiated (roughly half of the diodes were covered during irradiation). To address the first concern, we annealed the devices according to the ASTM standard [1]. The differences remained the same. To determine the irradiated area, we performed large area IBIC scans on several devices. Error! Reference source not found. below shows the IBIC maps of two devices one from each batch. The irradiated areas are approximately the same.

42 ENGINEERING↗
DOE OSTI2025

Hydrogen Adsorption, Reactivity, and Catalysis on Colloidal Iron Carbide Nanoparticles

Supported iron carbide particles have long served as catalysts for CO hydrogenation (the Fischer Tropsch synthesis, FTS) and continue to be attractive. Despite this, little is known about their chemistry. Reported here is a colloidal Fe x C nanoparticle (NP) model system that allows direct observation of surface hydrogen and CO, as well as quantification of the surface H. Dodecylamine-capped Fe x C NPs (DDA-Fe x C NPs) were synthesized through solution-phase carburization of Fe NPs and form stable colloids in low-polarity organic solvents. Treatment of these colloids with H 2 or D 2 produced highly hydrogenated materials, and FTIR spectra of DDA-Fe x C-D n showed that most of the D binds to carbides, with at least four distinct v(C-D) modes. The surface C-H(D) bonds were all reactive, transferring hydrogen to alkenes and other reagents in solution. Titration and ICP measurements showed a 0.17:1 ratio of added H:total Fe, or roughly. 40 H per 1.8 nm DDA-Fe x C NP. Conversely, CO was preferentially bound to surface Fe sites, with FTIR spectra showing a single broad v(CO) that shifted with CO coverage or co-adsorption of H 2 . The DDA-Fe x C NPs were active catalysts for both olefin hydrogenation and the FTS, under mild conditions and without catalyst pre-treatment. The CO hydrogenation reactions yielded a broad distribution of long-chain linear paraffins and olefins. Though quantitative comparisons with typical FTS results are not possible because of our use of sealed batch reactors and other factors, the observations of high catalytic reactivity demonstrate the relevance of this model system to iron-carbide catalysis. DFT calculations on model slab surfaces with varying iron carbide stoichiometries revealed that the thermodynamically preferred surface adsorption sites are C for H ads and Fe for CO ads . A variety of binding sites and binding energies were found for each adsorbate. We are unaware of previous studies indicating that a diverse array of C-H bonds is the primary source of reactive H on iron carbides. Experimentally, the diversity of *C-H sites was evident in reactions with H-atom donors and abstractors of different strengths, from both the reaction stoichiometries and IR spectra. The different surface–H binding energies correlate with the v(C-D) stretching frequencies. Furthermore, these insights into complex iron carbide surfaces and catalysis could assist catalyst design, and they showcase the importance of stoichiometric studies of reaction intermediates.

08 HYDROGEN↗
DOE OSTI2018

Purification of Zirconium Tetrachloride from UNF Cladding

The overarching objective of this project was to design and demonstrate (laboratory scale) a sublimation-based purification protocol to obtain pure zirconium chloride (ZrCl 4 ) from impure feeds derived from the direct chlorination of Zircaloy cladding materials. To achieve this objective the following goals were identified. (1) Design, construct and test an apparatus capable of subliming and purifying ZrCl 4 at room pressure and temperatures between 600 and 700K (sublimation point of ZrCl 4 606 K). (2) Investigate reactions that may be used to chemically alter the makeup of the impure ZrCl 4 feed materials to enhance their removal from ZrCl 4 in the sublimation-purification process. (3) Computationally model the reactions of metal chlorides in the gas phase under sublimation conditions (1 atm, 600 K) to determine if oligomers are formed and, if formed, how they would influence sublimation-based purification strategies. All experimental investigations performed as part of this project used stable (non-radioactive) compounds and materials as surrogates for the impure ZrCl 4 feedstocks that would be derived from the direct chlorination of Zircaloy claddings from spent nuclear fuel rods. No radiologically active materials were used at any time in this project. The impure ZrCl 4 feedstocks used in this study were derived from the direct chlorination of virgin Zircaloy claddings (Zirc-2, -4 and high Nb). The main goals of the project have been achieved during the three year period of this NEUP sponsored project. The impurities of greatest concern in the feedstocks were identified as the chloride salts of the alloying metals originally found in Zircaloy (Fe, Nb) and radiological contaminants (Cs, Sr, and Sb). A series of prototype sublimation apparatuses and heating arrangements were designed, constructed and tested. A general batch-sublimator design was identified and a 1.0 kg batch sublimator/oven system was fabricated from glass. Purification protocols developed on small scale runs were found to scale to 1.0 kg batches without problem. Purification factor ranges (PFs: wt% final / wt% initial ) for the impurity chlorides identified above have been measured: Fe (40 – 500); Nb (10 – 70); Cs (520 – 5000); Sr (250 – 1900); Sb (220 – 2300). Removal of iron is sensitive to the partial pressure of hydrogen (in nitrogen) flowing through the system. Lower PFs are obtained when the hydrogen partial pressure is lower. Lower flows are required to raise the iron PF. Removal of niobium is very sensitive to the endpoint determination of the forerun. When judged properly (color change yellow to white in sublimate) reasonable PFs (500) can be achieved. However, as discussed in the conclusions section, endpoint determination can be difficult and significantly lower PFs may be obtained when the endpoint is not correctly identified. Several methods of obtaining quantitative cesium, strontium and antimony PFs were investigated, the best being ICP-MS. The results reported here reflect very limited access to and ICP-MS instrument during the course of this project. It should be noted that, while significantly reduced amounts of antimony are observed in the product, antimony is also found in the exhaust trap after the condenser. Quantum chemical computational modeling studies of the reactions of metal chlorides in the gas phase provided important information about a chemical transport mechanism that could defeat sublimation-based purification strategies. Many metal chlorides are known to oligomerize in the gas phase to reduce their coordinative unsaturation. Dimerization of ZrCl 4 with iron or niobium chlorides would provide a mechanism by which both metal impurities could be carried over with the product during sublimation of the bulk material. Computational modeling conclusively showed that under sublimation conditions in the gas phase, dimerization reactions are not favored thermodynamically due to a significant entropic penalty for the reaction. Therefore, consistent with experimental observations, sublimation-based purification strategies remain viable. The experimental part of the report concludes with a short discussion of the issues that still remain in developing a technologically relevant Zircaloy recycling strategy. The final section of the report is a summary of the quantum chemical investigation of the reactions of metal chlorides in the gas phase under sublimation conditions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2017

Modular Caustic-Side Solvent Extraction Unit (MCU) Improved Solvent Quality and Processing Results - 17496

The Modular Caustic-Side Solvent Extraction (CSSX) Unit (MCU) is a process for Cs-137 removal from legacy nuclear waste. MCU began hot operations in 2008 with an initial planned operating life of three years and a design life of five years. After eight years of operation, MCU has processed more than 23.8 million liters (6.3 million gallons) of waste. The MCU process uses a solvent developed by Oak Ridge National Laboratory (ORNL) to extract the fission product Cs-137 from liquid salt waste, or salt solution (SS), stored in the Tank Farms at Savannah River Site (SRS) after the material has been treated through the Actinide Removal Process (ARP). Batches of SS are prepared and first fed to the ARP, where it is treated to become clarified SS (CSS), then fed to MCU. Once the Cs-137 has been extracted by the solvent, the CSS becomes decontaminated SS (DSS), which is ultimately processed into a cementitious grout at the Saltstone Production Facility (SPF) and subsequently disposed of at the Saltstone Disposal Facility (SDF). The Cs-137 is then stripped from the solvent into an aqueous strip effluent (SE) stream for incorporation into borosilicate glass at the Defense Waste Processing Facility (DWPF). The solvent used in the MCU process is made of four components: the extractant (extracts Cs-137 from CSS), the modifier (increases solubility of extractant and suppressor in the diluent), the suppressor (suppresses the effects of anions and improves stripping), and the diluent (controls density and viscosity). There have been two solvents utilized during the lifetime of MCU. The first solvent was developed by ORNL and was used to design the plant based on its cesium removal capability. Due to the extended life of MCU, in order to continue to minimize the curies disposed at SDF, a new solvent capable of removing more cesium from the CSS was developed by ORNL and implemented at MCU. The second MCU solvent, Next-Generation Solvent (NGS), uses the same modifier and diluent as the original solvent, but a different extractant and suppressor. Subsequent changes to the overall MCU flowsheet were made for the use of NGS. The success of cesium removal from the CSS is often presented as the cesium decontamination factor (DF) achieved for each batch processed. The DF is the ratio of the concentration of cesium in the CSS to the cesium concentration in the DSS. The BOBCalixC6 based solvent was capable of achieving DFs between 100 and 400 for each DSS batch. The NGS solvent is capable of achieving DFs greater than 40,000 for each DSS batch. The NGS blend solvent is currently meeting the goals of the pure NGS solvent. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Batched sparse direct solver design and evaluation in SuperLU_DIST

Over the course of interactions with various application teams, the need for batched sparse linear algebra functions has emerged in order to make more efficient use of the GPUs for many small and sparse linear algebra problems. In this paper, we present our recent work on a batched sparse direct solver for GPUs. The sparse LU factorization is computed by the levels of the elimination tree, leveraging the batched dense operations at each level and a new batched Scatter GPU kernel. The sparse triangular solve is computed by the level sets of the directed acyclic graph (DAG) of the triangular matrix. Batched operations overcome the large overhead associated with launching many small kernels. For medium sized matrix batches with not-so-small bandwidth, using an NVIDIA A100 GPU, our new batched sparse direct solver is orders of magnitude faster than a batched banded solver and uses less than one-tenth of the memory.

Boukaram, Wajih↗
DOE OSTI2019

Calibration and Statistical Performance of Al2O3: C Optically Stimulated Luminescent Dosimeters With and Without Annealing Using a 137Cs Source

A series of experiments were conducted using commercially available Al 2 O 3 :C optically stimulated luminescent dosimeters to provide a technical basis for their precise calibration and statistical performance at irradiated air kerma doses between 0.02 mGy and 5 mGy using 137 Cs. This study examines the dose response linearity, studies the background signal for annealed dosimeters, and compares the statistical performance of dosimeters that were annealed and not annealed prior to their irradiation and readout. The average and standard deviation for the response of groups of dosimeters annealed and nonannealed prior to their irradiation were determined at each delivered dose. The batch of dosimeters that were annealed prior to their irradiation exhibited a coefficient of variation in its mean dose response below 10% when using three or more irradiation trials at each delivered air kerma dose between 0.02 mGy and 5m Gy. The reader calibration factor was calculated using the response of the annealed batch of dosimeters and was determined to be 756 ± 7 photomultiplier tube counts perm Gy. Best estimates of the individual sensitivity factors were determined to be between 0.79 and 1.12 for the annealed batch of dosimeters. The minimum number of irradiations required to accurately determine the sensitivity factor of each individual dosimeter is reported with the recommended reader and dosimeter calibration procedures.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2017

Fate and transport of hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX) and its degradation products in sedimentary and volcanic rocks, Los Alamos, New Mexico

High-explosive compounds including hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX) were used extensively in weapons research and testing at Los Alamos National Laboratory (LANL). Liquid effluents containing RDX were released to an outfall pond that flowed to Cañon de Valle at LANL's Technical Area 16 (TA-16), resulting in the contamination of the alluvial, intermediate and regional groundwater bodies. Monitoring of groundwater within Cañon de Valle has shown persistent RDX in the intermediate perched zone located between 225 and 311 m below ground surface. Monitoring data also show detectable levels of RDX putative anaerobic degradation products. Batch and column experiments were conducted to determine the extent of adsorption-desorption and transport of RDX and its degradation products (MNX, DNX, and TNX) in major rock types that are within the RDX plume. All experiments in this paper were performed in the dark using water obtained from a well located at the center of the plume, which is fairly oxic and has a neutral pH of 7.5. Retardation factors and partitioning coefficient (K d ) values for RDX were calculated from batch experiments. Additionally, retardation factors and K d values for RDX and its degradation products were calibrated from column experiments using a one-dimensional transport model with equilibrium sorption (linear isotherm). Results from the column and batch experiments showed little to no sorption of RDX to the aquifer materials tested, with retardation factors ranging from 1.0 to 1.8 and K d values varying from 0 to 0.70 L/kg. Finally, results also showed no measurable differences between the transport properties of RDX and its degradation products.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2016

Fabrication of core-shell Fe{sub 3}O{sub 4}@MIL-100(Fe) magnetic microspheres for the removal of Cr(VI) in aqueous solution

Facile regeneration of an adsorbent is very important for commercial feasibility. One typical highly porous metal-organic framework (MOF) materials based on MIL-100(Fe) and magnetic iron oxide particles (denoted as MMCs) with diameter about of 350 nm were successfully synthesized. The growth of MIL-100(Fe) shell on the surface of Fe{sub 3}O{sub 4} was utilized precursor as crystal seed via in-situ step hydrothermal reaction. It is a simple way to obtain well organized core-shell MOF composites, compared to the step-by-step method. MMCs were firstly used to uptake of Cr(VI) anions in aqueous solution. Adsorption experiments were carried out in batch sorption mode investigating with the factors of contact time (0–1000 min), pH (from 2 to 12), dose of adsorbent (4–25 mg), and initial Cr(VI) concentration (range from 10 to 100 ppm). - Graphical abstract: One typical highly porous metal-organic framework (MOF) materials based on MIL-100(Fe) and magnetic iron oxide particles (denoted as MMCs) were successfully synthesized. Utilizing Fe{sub 3}O{sub 4} precursor as crystal seed to grow MIL-100(Fe) shell by in-situ step hydrothermal reaction. It is a simple way to obtain core-shell MOF composites. MMCs could effectively uptake of Cr(VI) anions in aqueous solution. - Highlights: • Fe{sub 3}O{sub 4}@MIL-100(Fe) composites with core-shell structure were successfully prepared through a simple method. • The influence factors on Cr(VI) adsorption by Fe{sub 3}O{sub 4}@MIL-100(Fe) were investigated. • Cr(VI) can efficiently adsorbed by Fe{sub 3}O{sub 4}@MIL-100(Fe) composites from aqueous solution.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Controls on Barite (BaSO 4 ) Precipitation in Unconventional Reservoirs

Barite (BaSO 4 ) precipitation is one of the most ubiquitous examples of secondary sulfate mineral scaling in shale oil and gas reservoirs. Often, a suite of chemical additives is used during fracturing operations to inhibit the accumulation of mineral scales, though their efficacy is widely varied and poorly understood. This study combines experimental data and multi-component numerical reactive transport modeling to offer a more comprehensive understanding of the geochemical behavior of barite accumulation in shale matrices under conditions typical of fracturing operations. A variety of additives and conditions are individually tested in batch reactor experiments to identify the factors controlling barite precipitation. Our experimental results demonstrate a pH dependence in the rate of barite precipitation, which we use to develop a predictive model including a pH-dependent term that satisfactorily reproduces our observations. Further, this model is then extended to consider the behavior of three major shale samples of highly variable mineralogy (Eagle Ford, Marcellus, and Barnett). This data-validated model offers a reliable tool to predict and ultimately mitigate against secondary mineral accumulation in unconventional shale reservoirs.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Accelerating Radiation Computations for Dynamical Models With Targeted Machine Learning and Code Optimization

Abstract Atmospheric radiation is the main driver of weather and climate, yet due to a complicated absorption spectrum, the precise treatment of radiative transfer in numerical weather and climate models is computationally unfeasible. Radiation parameterizations need to maximize computational efficiency as well as accuracy, and for predicting the future climate many greenhouse gases need to be included. In this work, neural networks (NNs) were developed to replace the gas optics computations in a modern radiation scheme (RTE+RRTMGP) by using carefully constructed models and training data. The NNs, implemented in Fortran and utilizing BLAS for batched inference, are faster by a factor of 1–6, depending on the software and hardware platforms. We combined the accelerated gas optics with a refactored radiative transfer solver, resulting in clear‐sky longwave (shortwave) fluxes being 3.5 (1.8) faster to compute on an Intel platform. The accuracy, evaluated with benchmark line‐by‐line computations across a large range of atmospheric conditions, is very similar to the original scheme with errors in heating rates and top‐of‐atmosphere radiative forcings typically below 0.1 K day −1 and 0.5 W m −2 , respectively. These results show that targeted machine learning, code restructuring techniques, and the use of numerical libraries can yield material gains in efficiency while retaining accuracy.

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

Modeling Thermal Interactions between Buildings in an Urban Context

Thermal interactions through longwave radiation exchange between buildings, especially in a dense urban environment, can strongly influence a building’s energy use and environmental impact. However, these interactions are either neglected or oversimplified in urban building energy modeling. We developed a new feature in EnergyPlus to explicitly consider this term in the surface heat balance calculations and developed an algorithm to batch calculating the surrounding surfaces’ view factors using a ray-tracing technique. We conducted a case study with a district in the Chicago downtown area to evaluate the longwave radiant heat exchange effects between urban buildings. Results show that the impact of the longwave radiant effects on annual energy use ranges from 0.1% to 3.3% increase for cooling and 0.3% to 3.6% decrease for heating, varying among individual buildings. At the district level, the total energy demand increases by 1.39% for cooling and decreases 0.45% for heating. We also observe the longwave radiation can increase the exterior surface temperature by up to 10 °C for certain exterior surfaces. These findings justify a detailed and accurate way to consider the thermal interactions between buildings in an urban context to inform urban planning and design.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗