Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “minimizing costs”

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 991 records · Page 55
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Enhanced 2'-Fucosyllactose production by engineered Saccharomyces cerevisiae using xylose as a co-substrate

2'-Fucosyllactose (2'-FL), a human milk oligosaccharide with confirmed benefits for infant health, is a promising infant formula ingredient. Although Escherichia coli, Saccharomyces cerevisiae, Corynebacterium glutamicum, and Bacillus subtilis have been engineered to produce 2'-FL, their titers and productivities need be improved for economic production. Glucose along with lactose have been used as substrates for producing 2'-FL, but accu- mulation of by-products due to overflow metabolism of glucose hampered efficient production of 2' -FL regardless of a host strain. To circumvent this problem, we used xylose, which is the second most abundant sugar in plant cell wall hydrolysates and is metabolized through oxidative metabolism, for the production of 2'-FL by engi- neered yeast. Specifically, we modified an engineered S. cerevisiae strain capable of assimilating xylose to pro- duce 2' -FL from a mixture of xylose and lactose. First, a lactose transporter (Lac12) from Kluyveromyces lactis was introduced. Second, a heterologous 2'-FL biosynthetic pathway consisting of enzymes Gmd, WcaG, and WbgL from Escherichia coli was introduced. Third, we adjusted expression levels of the heterologous genes to maximize 2'-FL production. The resulting engineered yeast produced 25.5 g/L of 2'-FL with a volumetric productivity of 0.35 g/L∙h in a fed-batch fermentation with lactose and xylose feeding to mitigate the glucose repression. Interestingly, the major location of produced 2'-FL by the engineered yeast can be changed using different culture media. While 72% of the produced 2'-FL was secreted when a complex medium was used, 82% of the produced 2'-FL remained inside the cells when a minimal medium was used. In conclusion, as yeast extract is already used as food and animal feed ingredients, 2'-FL enriched yeast extract can be produced cost-effectively using the 2'-FL-accumu- lating yeast cells.

2′ -fucosyllactose↗
DOE OSTI2025

Decoupling electrode kinetics to elucidate reaction mechanisms in alkaline water electrolysis

Alkaline water electrolysis (AWE) presents key advantages, including reduced material costs, enhanced operational stability, and compatibility with non-precious metal catalysts, positioning it as a scalable route for hydrogen production. In this study, we introduce a minimally invasive single-cell configuration incorporating a reference electrode via diaphragm extension to form an internal ion channel. This setup, combined with an interfaced potentiostat and auxiliary electrometer, enables real-time, independent monitoring of anode and cathode behavior, offering high-resolution electrochemical diagnostics. While it is well established that the hydrogen evolution reaction (HER) exhibits sluggish kinetics in alkaline media, our study reveals that this limitation persists even in practical AWE systems where nickel-based substrates are used as electrodes. This observation is supported by both experimental data and voltage breakdown modeling. Arrhenius-type analysis reveals that localized electric fields induced by catalysts shift the reaction kinetics from classical Butler–Volmer behavior toward a Marcus-like regime, where interfacial molecular dynamics and bimolecular charge transfer dominate. We propose a semi-empirical model and a surficial reaction mechanism to describe these dynamics. This work underscores the critical need for cathode innovation and provides a rational framework for designing advanced catalysts and electrode architectures to optimize AWE performance.

08 HYDROGEN↗
DOE OSTI2023

Contribution of carbon pricing to meeting a mid-century net zero target

A mid-century net zero target creates a challenge for reducing the emissions of emissions-intensive, trade-exposed sectors with high cost mitigation options. These sectors include aluminium, cement, chemicals, iron and steel, lime, pulp and paper and petroleum refining. Available studies agree that decarbonization of these sectors is possible by mid-century if more ambitious policies are implemented soon. Existing carbon pricing policies have had limited impact on the emissions of these sectors because their marginal abatement costs almost always exceed the tax rate or allowance price. But emissions trading systems with free allowance allocations to emissions-intensive, trade-exposed sectors have minimized the adverse economic impacts and associated leakage. Internationally coordinated policies are unlikely, so implementing more ambitious policies creates a risk of leakage. This paper presents policy packages a country can implement to accelerate emission reduction by these sectors with minimal risk of leakage. To comply with international trade law the policy packages differ for producers whose goods compete with imports in the domestic market and producers whose goods are exported. Carbon pricing is a critical component of each package due its ability to minimize the risk of adverse economic impacts on domestic industry, support innovation and generate revenue. The revenue can be used to assist groups adversely impacted by the domestic price and production changes due to carbon pricing and to build public support for the policies.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Effects of Media Nutrient Variation on Microalgae Productivity and Economics During Semi-Continuous Cultivation

The development of large-scale microalgae growth for biofuel production is currently limited by the cost of biomass production. However, new approaches to infrastructure and cultivation practices are bringing the field closer to realization. Macronutrients in the cultivation media contribute significant costs, especially since their concentrations have not been optimized for specific strains and conditions. Environmental photobioreactors (ePBRs) were used to simulate cultivation under outdoor conditions, during which the nitrogen and phosphorus levels in the media were varied. The growth of two potential biofuel production strains, Picochlorum celeri and Tetraselmis striata, with varying nutrient inputs during summer and winter scripts, respectively, was studied. This study demonstrated that nitrogen and phosphorus in f/2 media could be reduced by more than 60% from the standard formulation, while maintaining growth rates in a semi-continuous harvesting approach. Experiments comparing the standard and reduced nutrient input concentrations were also conducted for both species in 820 L outdoor raceway ponds, in Mesa, AZ. P. celeri grown in these ponds in October had a growth rate of 10.6 ± 0.7 g/m 2 /day and 10.6 ± 0.3 g/m 2 /day for the standard and low-nutrient P. celeri ponds, respectively. T. striata grown in April–May had a growth rate of 16.6 ± 1.4 g/m 2 /day for the standard nutrient input ponds and 17.4 ± 1.1 g/m 2 /day for the low-nutrient input ponds, and in October 14.5 ± 0.6 g/m 2 /day for standard nutrient ponds and 14.4 ± 0.6 g/m 2 /day for low-nutrient ponds. These outdoor data therefore confirmed the indoor ePBR data. Techno-economic analysis shows that, if high growth rates can be attained at lower nutrient concentrations, a reduction of at least 60% in nutrient costs can be achieved. Such results highlight the importance of managing macronutrient media inputs, as these have a considerable contribution to biomass production costs in large-scale facilities. The analysis also points to the importance of maintaining high spent medium recycling rates in an industrial deployment, so as to minimize the losses of nitrogen and phosphorus compounds.

09 BIOMASS FUELS↗
DOE OSTI2024

Cost-efficient temperature mediation of DPMS synthesis in flow using continuous flow chemistry

An efficient synthesis of diphenylmethylsilanol (DPMS) utilizing continuous flow chemistry was performed using a coil reactor and an ice bath. The inclusion of careful temperature control minimized the exothermic production of an unwanted dimer. Yields of up to 88% were obtained and dimer formation as low as 2.2% was confirmed by NMR. This material will be used in the formulation to make SX358 and its physical and chemical characteristics will be determined using DSC, DMA, TMA, compression set testing and FTIR.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2026

Progressing Analysis of Variable Electric Rates (PAVER) Study

The Progressing Analysis of Variable Electric Rates (PAVER) study analyzed the impact of a range of time-varying electric rates on the performance of a regional electric grid and the resulting costs for participating and non-participating customers. This analysis leveraged and extended the work of PNNL’s Distribution System Operator with Transactive (DSO+T) study. Five different rate designs were included: a flat volumetric energy charge, a typical Time of Use (TOU) rate, a dynamic energy (DE) rate (based on wholesale locational marginal prices), a dynamic energy and capacity (DE+C) rate, and, finally, a Block and Swing (B&S) rate that billed customers based on their average load profile at constant pricing, but used the DE+C dynamic price for load deviations from their average profile. These rates were analyzed in a large-scale co-simulation of an entire regional grid with a customer population representative of the current state. A large fraction (80%) of residential and commercial customers were assumed to participate in these time-varying rates with automatically controlled HVAC, water heaters, electric vehicles, and batteries. This study assumed no industrial sector participation. The DE and DE+C rates saw system peak loads reduced by 6-7%, while the large participation in the TOU rate case saw a significant rebound effect and a resulting peak load increase of >5%. The impacts to the annual and peak system demand impacted system wholesale prices and the overall grid operating costs. This cost structure determined the revenue needed to be collected from customers by each rate design. Participating customers on the DE and DE+C rates (located in one of the modeled DSOs) saw reductions in average annual electricity bills of 11-17% with average increases in monthly bill variation of no more than 13%. At such high participation levels, TOU customers saw 10% higher average annual bills (due to system-wide rebound effects) and average increased monthly bill variation of 16%. Residential owners of large flexible loads (such as electric vehicles) saw larger bill savings (17-20%) when on a fully dynamic rate. The presence of on-site generation (such as rooftop solar) did not appear to appreciably change customer outcomes. Customers on the Block and Swing rate did see 6% lower monthly bill variation (as intended) than the flat rate case, but at the expense of appreciable bill savings, which were only 3%, comparable to the savings seen by non-participants. Given this finding we recommend that additional research be conducted into how best various bill protection mechanisms can balance minimizing customer bill variation with providing financial incentives commensurate with the flexibility customers provide. We also recommend that customer outcomes be explored across a range of regions using current actual customer and system cost data.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2023

Cummins R-SOFC System Development

The overall purpose of this project was to reduce the Reversible-Solid Oxide Fuel Cell (R-SOFC) system cost by developing two technologies, an improved cell design and the incorporation of an ejector in the fuel recycle loop instead of a blower. A Simulink model of the baseline SOFC system was developed and calibrated with experimental test data. The R-SOFC system model was built by integrating GT Suite developed models of the steam generation components into the baseline Simulink SOFC system model. The ability to run the stack in SOEC operating mode was also added to the model. The system model was used to explore the ability of the R-SOFC system to meet operational constraints on Steam/Carbon ratio and H2 concentration on the fuel side electrode. A CFD ejector model was developed and used to explore a range of ejector design parameters, leading to the final ejector design that was prototyped for testing. A prototype steam ejector was first tested in a laboratory environment using room temperature air. The steam ejector was subsequently tested using the full hot recycle loop with all relevant heat exchangers and steam generation components. The test conditions utilized temperatures, pressures, and flow rates expected in an R-SOFC application. Throughout the experimental testing work, ejector performance test data was used to improve and then validate the CFD ejector model. A CFD cell model was developed and used to optimize thermal gradients, voltage, and cost of a new cell substrate design. A CFD comparison of co-flow and cross-flow cell designs informed the decision to use a co-flow design for the new cell substrate. Multiple rounds of CFD simulation were used to improve the cell design to minimize the variation in air and fuel distribution across cell channels and to minimize the variation in air and fuel distribution across different cells in the stack. A few prototypes of the new cell substrate design were produced and validated in a laboratory environment by thermally spraying and verifying that they met established manufacturing specifications. The cell manufacturing process was adjusted in order to bring these metrics within acceptable tolerances. Cummins’ internal calculations show that the new cell design reduces cost ~50% compared to the baseline cell, while the ejector + superheater/boiler concept reduces cost of the recycle loop by ~40%. The impact of these cost reductions on the cost of producing H2 will depend on the specific system where they are applied. Therefore, a Techno-Economic Analysis was completed using system cost as a variable, and showing how the NREL Current and Future system costs translate into H2 production cost.

Henrichsen, Lars↗
DOE OSTI2020

Improved Value of the Gasoline and Fuel Oil Co-Product Fractions Generated by the PNNL/LanzaTech Alcohol-to-Jet Process (Final Report)

The BETO-funded PNNL collaboration with LanzaTech, Inc. utilizes patented technology to address the need for low aromatic jet fuel blendstocks. The PNNL/LanzaTech alcohol-to-jet (ATJ) process converts ethanol from LanzaTech’s syngas fermentation process to jet-range isoparaffins. The aim of this Cooperative Research Development Agreement (CRADA) project was to improve the commercial viability of the LanzaTech/ PNNL ATJ process by developing two new co-product options. Two Focus Areas (FA) were explored: FA1) Increasing the RON of the lighter-than-jet gasoline fraction above 98; and FA2) Creating a synthetic lubricant base oil from the heavier-than-jet fraction. For FA 1, multiple processing approaches were evaluated, and RONs ranging from 96-100 were obtained, each comprising different processing costs. The most cost effective approach identified was a single-step process and yielded a liquid product of 97. Thus, additional development is required to obtain a RON > 98. We will continue work on this focus area with separate Direct Funding Opportunity (DFO) funding within the DOE-BETO Co-Optima Consortium (expected to begin Q2-FY21). Experimental work will focus on i) increasing single pass conversion to > 50%, and ii) obtaining a > 98 RON product. Technoeconomic analysis performed in this project suggests cost competitiveness provided these technical targets can be met. For FA2, multiple approaches were evaluated for increasing the viscosity index (VI) of the heavier-than-jet fraction to at least 120 in order to meet ASTM specifications for a Group III Base Oil (ASTM D6074). All other ASTM specifications are already met. A viscosity index of 115 was obtained, and with minimal undesirable cracking products. We believe a VI of 115 is about the highest possible from a catalytic approach, per review of the patent literature. In order to further increase the VI to > 120 we believe additives and/or liquid-liquid extraction is required. TEA performed earlier in the project suggests cost competitiveness with market prices assuming cracking byproducts are kept < 10 wt.%. If TEA updated with this additional processing still projects economic feasibility next steps could be taken in a subsequent effort.

02 PETROLEUM↗
DOE OSTI2025

SIDDA: SInkhorn Dynamic Domain Adaptation

Modern neural networks (NNs) often do not generalize well in the presence of a "covariate shift"; that is, in situations where the training and test data distributions differ, but the conditional distribution of classification labels remains unchanged. In such cases, NN generalization can be reduced to a problem of learning more domain-invariant features. Domain adaptation (DA) methods include a range of techniques aimed at achieving this; however, these methods have struggled with the need for extensive hyperparameter tuning, which then incurs significant computational costs. In this work, we introduce SIDDA, an out-of-the-box DA training algorithm built upon the Sinkhorn divergence, that can achieve effective domain alignment with minimal hyperparameter tuning and computational overhead. We demonstrate the efficacy of our method on multiple simulated and real datasets of varying complexity, including simple shapes, handwritten digits, and real astronomical observations. SIDDA is compatible with a variety of NN architectures, and it works particularly well in improving classification accuracy and model calibration when paired with equivariant neural networks (ENNs). We find that SIDDA enhances the generalization capabilities of NNs, achieving up to a ≈40% improvement in classification accuracy on unlabeled target data. We also study the efficacy of DA on ENNs with respect to the varying group orders of the dihedral group DN, and find that the model performance improves as the degree of equivariance increases. Finally, we find that SIDDA enhances model calibration on both source and target data--achieving over an order of magnitude improvement in the ECE and Brier score. SIDDA's versatility, combined with its automated approach to domain alignment, has the potential to advance multi-dataset studies by enabling the development of highly generalizable models.

Pandya, Sneh [Northeastern U.]↗
DOE OSTI2021

Land-based wind turbines with flexible rail-transportable blades – Part 1: Conceptual design and aeroservoelastic performance

This work investigates the conceptual design and the aeroservoelastic performance of land-based wind turbines whose blades can be transported on rail via controlled bending. The turbines have a nameplate power of 5 MW and a rotor diameter of 206 m, and they aim to represent the next generation of land-based machines. Three upwind designs and two downwind designs are presented, combining different design goals together with conventional glass and pultruded carbon fiber laminates in the spar caps. One of the five blade designs is segmented and serves as a benchmark to the state of the art in industry. The results show that controlled flexing requires a reduction in the flapwise stiffness of the blades, but it represents a promising pathway for increasing the size of land-based wind turbine rotors. Given the required stiffness, the rotor can be designed either downwind with standard rotor preconing and nacelle uptilt angles or upwind with higher-than-usual angles. A downwind-specific controller is also presented, featuring a cut-out wind speed reduced to 19 m s –1 and a pitch-to-stall shutdown strategy to minimize blade tip deflections toward the tower. The flexible upwind and downwind rotor designs equipped with pultruded carbon fiber spar caps are found to generate the lowest levelized cost of energy, 2.9 % and 1.3 %, respectively, less than the segmented design. The paper concludes with several recommendations for future work in the area of large flexible wind turbine rotors.

17 WIND ENERGY↗
DOE OSTI2023

Freeform thermoelectrics in single-step manufacturing: additive manufacturing of bismuth-telluride thermoelectrics

The project succeeded in producing crack-free Bismuth Telluride thermoelectric parts with density exceeding 98% through laser powder bed fusion (LPBF) additive manufacturing (AM). This greatly exceeded the highest previously reported density of 88% and is the highest among all semiconducting materials processed by LPBF. The additively manufactured material shows comparable Seebeck coefficient as conventional form and can be made into complex geometries with reduced material loss. On the other hand, measured properties are dramatically sensitive to the AM process parameters used, such that with identical composition, the Seebeck coefficient can be controllably tuned from +120 µV/K to -207 µV/K, which means the material switches between an n-type to a p-type semiconductor depending on processing. These changes are accompanied by significant differences in the as-processed microstructure due to rapid solidification. A machine learning protocol was developed and greatly reduced the experimental burden of the project, reducing the typical process optimization period of 2 years to 6 months. The project was fully successful in the objective of producing defect-free, complex geometry of bismuth-telluride parts through LPBF, but only partially successful in achieving performance goals. First, cost reduction of manufacturing, as measured by material waste, was successfully reduced by up to 70% compared to conventional manufacturing methods. This exceeded the proposed 30% reduction in materials waste needed to reach the 20% cost reduction goal of the project. On the other hand, the device performance, as measured by Seebeck coefficient, failed to reach the 40% improvement in efficiency. Rather, we observe comparable Seebeck coefficient between AM samples and conventionally processed counterparts. The device-level efficiency improvement does exceed 40% for complex geometry samples due to shape-induced increase in temperature gradients, but this was not the originally proposed metric. The machine learning approach developed in this project greatly accelerated the process optimization and can be adopted for fast development of AM processing parameters for other brittle and otherwise difficult-to-print materials. For the public, we deliver an efficient and widely adoptable process for incorporating waste-heat harvesting thermoelectric devices in both industrial and commercial heat exchangers. The geometric flexibility allows the capturing device to conform to the shape of the heat source to improve the system-level conversion efficiency. The technique can be deployed on any commercial LBPF systems with zero modifications, thus poses minimal adoption barrier for any manufacturer that already employed AM technology. Beyond bismuth-telluride, the machine-learning guided optimization protocol can be used in the future to reduce both the time and cost of process development for AM of other energy conversion and harvesting materials.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Direct flue gas capture for algae cultivation and subsequent valorization: evaluating life cycle emissions and costs

Algae cultivation and processing is an important pathway under discussion within the broader CO 2 capture and utilization umbrella. Here, we discuss the results of a life-cycle analysis and techno-economic analysis of a pilot-scale photobioreactor that uses flue gas directly from natural gas or biogas combustion at 3–5% CO 2 concentration. The system requires minimal freshwater use as it has been successfully run with industrial wastewater and has a much smaller areal footprint compared with open pond cultivation. Introducing the flue gas directly to the photobioreactor avoids the need for CO 2 separation and pressurization, which is undertaken in many other algae cultivation systems. For the end-use of the biomass, the default case assumes conversion of algae to liquid fuels via hydrothermal liquefaction. The results indicate that the pilot-scale system has a higher cost, and comparable greenhouse gas emissions compared to pond-based systems, especially as the grid is anticipated to evolve to a lower carbon intensity. The costs of algae biofuel production range from $\$12–16$ per GGE at the current pilot scale. Depending on whether the source of the carbon is fossil or biogenic, the net emissions are 68 g CO 2 e per MJ and −4 g CO 2 e per MJ respectively. If the marine algae species is used instead of the freshwater species, it offers an additional 16 g CO 2 e per MJ carbon fixation in the form of calcium carbonate. The findings point to broadly desirable trends in GHG emissions and costs, while the discussion aims to shed light on areas that could further improve the scalability of the system.

20 FOSSIL-FUELED POWER PLANTS↗
DOE OSTI2026

Data for An End-to-End Pipeline for Succinic Acid Production at an Industrially Relevant Scale Using Issatchenkia orientalis

Microbial production of succinic acid (SA) at an industrially relevant scale has been hindered by high downstream processing costs arising from neutral pH fermentation for over three decades. Here, we metabolically engineer the acid-tolerant yeast Issatchenkia orientalis for SA production, attaining the highest titers in sugar-based media at low pH (pH 3) in fed-batch fermentations, i.e. 109.5 g/L in minimal medium and 104.6 g/L in sugarcane juice medium. We further perform batch fermentation using sugarcane juice medium in a pilot-scale fermenter (300×) and achieve 63.1 g/L of SA, which can be directly crystallized with a yield of 64.0%. Finally, we simulate an end-to-end low-pH SA production pipeline, and techno-economic analysis and life cycle assessment indicate our process is financially viable and can reduce greenhouse gas emissions by 34–90% relative to fossil-based production processes. We expect I. orientalis can serve as a general industrial platform for production of organic acids.

Metabolomics↗
DOE OSTI2024

Leverage modern artificial intelligence (AI) enabled systems for waste reduction

Manufacturing industries continue to face challenges in reducing waste, as upstream strategies such as source reduction and product redesign require a deeper understanding of processes compared to conventional recycling methods. Recent advancements in artificial intelligence (AI) and machine learning (ML) have opened new opportunities to integrate modern computational techniques with traditional waste minimization strategies. This paper explores AI-enabled approaches for product redesign, source reduction, and recycling that can significantly reduce waste generation while improving efficiency and sustainability. AI-driven material substitution and lightweighting in product design enable discovery of novel materials with optimized properties, reducing waste without compromising performance. Reinforcement learning models optimize process parameters, raw material specifications, and machine sequencing to minimize production losses, while Industrial Internet of Things (IIoT) systems paired with AI analytics enhance real-time waste tracking, predictive maintenance, and quality inspection. Furthermore, AI-based demand forecasting and production planning reduce overproduction and excess inventory, as demonstrated in industrial applications. In recycling, ML-powered pattern recognition and robotic sorting technologies achieve higher accuracy in waste segregation, directly improving recycling efficiency. Complementary solutions such as smart bins and AI-enabled waste pickup scheduling optimize collection logistics, reducing both costs and emissions. Although implementation requires upfront investment in infrastructure and training, the long-term benefits include higher material efficiency, reduced waste, improved product quality, and stronger sustainability outcomes across the supply chain. By leveraging AI-enabled systems, manufacturers can align waste minimization efforts with circular economy principles, creating scalable solutions for both industry and society.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

Advanced one-pot deconstruction and valorization of lignocellulosic biomass into triacetic acid lactone using Rhodosporidium toruloides

Abstract Background Rhodosporidium toruloides is capable of co-utilization of complex carbon sources and robust growth from lignocellulosic hydrolysates. This oleaginous yeast is therefore an attractive host for heterologous production of valuable bioproducts at high titers from low-cost, deconstructed biomass in an economically and environmentally sustainable manner. Here we demonstrate this by engineering R. toruloides to produce the polyketide triacetic acid lactone (TAL) directly from unfiltered hydrolysate deconstructed from biomass with minimal unit process operations. Results Introduction of the 2-pyrone synthase gene into R. toruloides enabled the organism to produce 2.4 g/L TAL from simple media or 2.0 g/L from hydrolysate produced from sorghum biomass. Both of these titers are on par with titers from other better-studied microbial hosts after they had been heavily engineered. We next demonstrate that filtered hydrolysates produced from ensiled sorghum are superior to those derived from dried sorghum for TAL production, likely due to the substantial organic acids produced during ensiling. We also demonstrate that the organic acids found in ensiled biomass can be used for direct synthesis of ionic liquids within the biomass pretreatment process, enabling consolidation of unit operations of in-situ ionic liquid synthesis, pretreatment, saccharification, and fermentation into a one-pot, separations-free process. Finally, we demonstrate this consolidation in a 2 L bioreactor using unfiltered hydrolysate, producing 3.9 g/L TAL. Conclusion Many steps involved in deconstructing biomass into fermentable substrate can be combined into a distinct operation, and directly fed to cultures of engineered R. toruloides cultures for subsequent valorization into gram per liter titers of TAL in a cost-effective manner.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Geometry-aware framework for deep energy method: An application to structural mechanics with hyperelastic materials

Here, in this work, we introduce a novel physics-informed framework named the Geometry-Aware Deep Energy Method (GADEM) for solving structural mechanics problems on different geometries. As the weak form of the physical system equation (or the energy-based approach) has demonstrated clear advantages compared to the strong form for solving solid mechanics problems, GADEM employs the weak form and aims to infer the solution on multiple shapes of geometries. Integrating a geometry-aware framework into an energy-based method results in an effective physics-informed deep learning model in terms of accuracy and computational cost. Different ways to represent the geometric information and to encode the geometric latent vectors are investigated in this work. We introduce a loss function of GADEM which is minimized based on the potential energy of all considered geometries. An adaptive learning method is also employed for the sampling of collocation points to enhance the performance of GADEM. We present some applications of GADEM to solve solid mechanics problems, including a loading simulation of a toy tire involving contact mechanics and large deformation hyperelasticity. The numerical results of this work demonstrate the remarkable capability of GADEM to infer the solution on various and new shapes of geometries using only one trained model.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Approximating a linear multiplicative objective in watershed management optimization

Implementing management practices in a cost-efficient manner is critical for regional efforts to reduce the amount of pollutants entering the Chesapeake Bay. We study the problem of selecting a subset of practices that minimizes pollutant load—subject to budgetary and environmental constraints—as simulated in a widely used regulatory watershed model. Mimicking the computation of pollutant load in the regulatory model, we formulate this problem as a continuous optimization model with a linear multiplicative objective function and linear constraints. To lay the groundwork for incorporating additional stakeholder requirements in the future, especially those that would require integer variables, we present and study a continuous linear optimization model that approximates the nonlinear model. The linear model, which requires an exponential number of variables, arises naturally as an alternative model for the same underlying physical process. We examine the theoretical behavior of these optimization models and investigate restrictions of the linear model to handle its large number of variables. Through extensive computational tests on real and randomly generated instances, we demonstrate that the linear model and its restrictions provide optimal solutions close to those of the nonlinear model in practice, despite poor approximation properties in the worst case. We conclude that the linear model—together with our approach to handling its large number of variables—provides a viable framework from which to extend the optimization model to better meet the needs of the Chesapeake Bay watershed management stakeholders.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

A scalable algorithm for the optimization of neural network architectures

In this work, we propose a new scalable method to optimize the architecture of an artificial neural network. The proposed algorithm, called Greedy Search for Neural Network Architecture, aims to determine a neural network with minimal number of layers that is at least as performant as neural networks of the same structure identified by other hyperparameter search algorithms in terms of accuracy and computational cost. Numerical results performed on benchmark datasets show that, for these datasets, our method outperforms state-of-the-art hyperparameter optimization algorithms in terms of attainable predictive performance by the selected neural network architecture, and time-to-solution for the hyperparameter optimization to complete.

97 MATHEMATICS AND COMPUTING↗