Search NASA⌕ Search

SEARCH · Search NASA

Results for “underperformances”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Assessing geothermal/solar hybridization – Integrating a solar thermal topping cycle into a geothermal bottoming cycle with energy storage

Geothermal and concentrating solar power (CSP) technologies typically use heat at different temperatures in their commercial deployments. This enables a technically viable hybridization of a solar topping cycle and a geothermal bottoming cycle at locations where both resources are available. In this article, an underperforming geothermal power system based at Burley, Idaho is used as a baseline to investigate the technical and economic potential of such a hybrid cycle. A direct thermal energy storage (TES) system is also integrated to overcome the intermittency of the solar resource. Design and off-design behavior of key components are modelled to simulate the annual performance of the hybrid system on an hourly basis. The sizing of the solar field and thermal storage is investigated and is seen to significantly impact the annual electricity generation and efficiency. Various cost scenarios for the solar field and thermal energy storage are investigated. An economic metric - levelized cost of electricity (LCOE) - is used to optimize the solar field sizing and TES capacity. The hybrid plant converts the additional solar heat input into additional work with an efficiency of 32.9%. Retrofitting a geothermal plant with solar and eight hours of energy storage can achieve an LCOE of 0.136 $/kWhe in current cost scenarios and 0.081 $/kWhe in a future cost reduction scenario. Although the hybrid system cannot directly compete with current PV systems without batteries, it has an LCOE 32% lower than that of a PV-battery system. This paper provides insights into the research and development of the future grid with a high renewable energy penetration and encourages further study of energy hybridization for improved efficiencies and economics.

15 GEOTHERMAL ENERGY↗
DOE OSTI2025

Physics-based hybrid machine learning for critical heat flux prediction with uncertainty quantification

Critical heat flux (CHF) is a key quantity in nuclear system modeling due to its impact on heat transfer, safety margins, and reactor performance. This study develops and validates an uncertainty-aware hybrid modeling approach that combines machine learning with physics-based models to predict CHF in cases of dryout. The Biasi and Bowring empirical correlations were paired with three ML uncertainty quantification (UQ) techniques: deep neural network (DNN) ensembles, Bayesian neural networks (BNNs), and deep Gaussian processes (DGPs). A pure ML model without a base model was evaluated for comparison. Model performance was assessed under plentiful (7,350 points) and limited (9 points) training data scenarios using parity, uncertainty distributions, and calibration curves. Results show that the Biasi hybrid DNN ensemble achieved the best overall performance, with a mean absolute relative error of 1.846%, and well-calibrated uncertainty estimates. The BNN-based hybrids showed slightly higher error (2.14%) but superior uncertainty calibration. DGP models underperformed, with over 6% error and poor uncertainty calibration. All hybrid models outperformed pure machine learning configurations, demonstrating resistance against data scarcity. These findings indicate that hybrid modeling significantly improves predictive accuracy, interpretability, and resilience to data scarcity. The integration of uncertainty awareness provides actionable confidence in CHF predictions, which is vital for safety-critical decisions in nuclear applications. This hybrid approach offers a viable pathway for deploying ML models in reactor analysis tools while preserving domain knowledge and physical consistency.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2023

Biomass pretreatment with distillable ionic liquids for an effective recycling and recovery approach

Ionic liquid (IL) pretreatment methods show incredible promise for the efficient conversion of lignocellulosic feedstocks to fuels and chemicals. Given their low vapor pressures, distillation-based methods of extracting ionic liquids out of biomass post-pretreatment have historically been ignored in favor of alternative methods. We demonstrate a process to distill four acetate-based ionic liquids ([EthA][OAc], [PropA][OAc], [MAEthA][OAc], and [DMAEthA][OAc]) at low pressure and high purity that overcome some disadvantages of “water washing” and “one pot” recovery methods. Out of four tested ILs, ethanolamine acetate ([EthA][OAc]) is shown to have the most agreeable conversion metrics for commercial bioconversion processes achieving 73.6% and 51.4% of theoretical glucose and xylose yields respectively and >85 % recovery rates. Our process metrics are factored into a techno-economic analysis where [EthA][OAc] distillation is compared to other recovery methods as well as ethanolamine pretreatment at both milliliter and liter scales. In conclusion, although our TEA shows [EthA][OAc] distillation underperforming against other processes, we show a step-by-step avenue to reduce sugar production cost below the wholesale dextrose price at scale.

09 BIOMASS FUELS↗
DOE OSTI2025

Influence of extreme temperature conditions on CO 2 direct air capture using amino-acid solutions

Geological features play a pivotal role in determining the feasibility of deploying CO₂ direct air capture (DAC) technologies, primarily because they influence the availability of cost-effective energy sources, such as natural gas and geothermal energy, and also due to the potential for CO₂ sequestration. Many regions face challenges due to variable weather conditions including seasonal temperature fluctuations, high or low humidity, and sub-ambient temperatures. These extremes can reduce DAC performance or even lead to catastrophic events. Aqueous solvents considered for DAC systems are particularly vulnerable to seasonal variations in colder climates, where the solvent may underperform or freeze. It is therefore essential to investigate the CO₂ capture efficiency of aqueous solvents across a broad range of environmental temperatures, spanning sub-zero to hot conditions (>30 °C). In this study, DAC operation is examined using a high-flux solvent–air crossflow contactor under two major weather scenarios: (i) cold conditions below 0 °C and (ii) hot conditions above 30 °C. A parametric study is conducted to investigate the contactor performance regarding CO₂ removal efficiency, uptake capacity, and reaction kinetics versus temperature when the air velocity through the contactor exceeds 1 m/s. The efficacy of the contactor is systematically investigated using various anti-freeze amino-acid solvent formulations. A mass-transfer mechanistic model is developed to assess the process performance over a wide temperature range and propose scalable design guidelines. Machine learning is also employed to identify key parameters affecting the CO₂ capture efficiency. It is shown that air velocity and temperature are the primary factors influencing CO₂ uptake. Based on performance data obtained under subfreezing temperatures, a technoeconomic analysis is conducted to evaluate the feasibility of using aqueous solvents in seasonal cold regions. In conclusion, the findings of this study provide valuable insights into siting considerations for deploying solvent-based DAC, thereby contributing to the advancement of sustainable carbon removal solutions.

Air–liquid contactor↗
DOE OSTI2026

Robust optimization of flexible diafiltration systems for critical mineral separations

This paper provides major contributions in expanding the literature for membrane process design with critical mineral recovery applications and showcasing the importance of robust design techniques for reducing risks of underperformance in such systems. Here, a membrane process flowsheet featuring PrOMMiS membrane models for recovering lithium/cobalt from spent batteries is showcased, uncertainty in membrane sieving and localized fouling are considered, and robust designs are obtained using the PyROS toolset. This paper is intended for a general audience of researchers working in critical minerals, membranes, and optimization related areas.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Persistent fatty acid catabolism during plant oil synthesis

Plant lipids are an essential energy source for diets and are a sustainable alternative to petroleum-based fuels and feedstocks. Fatty acid breakdown during seed germination is crucial for seedling establishment but unexpected during seed filling. Here, we demonstrate that the simultaneous biosynthesis and degradation of fatty acids begins early and continues across all phases of oil filling and throughout the photoperiod. Tests in camelina, rapeseed, and an engineered high-oil tobacco line confirmed that concomitant synthesis and breakdown in oil-producing tissues over development is the rule rather than the exception. Furthermore, we show that transgenics, designed to elevate fatty acid biosynthesis, failed to achieve anticipated increases in storage lipid levels due to increased degradation, potentially explaining the underperformance of engineered lines compared to expectations more generally.

coenzyme A↗
DOE OSTI2023

Long-term carbon intensity reduction potential of K-12 school buildings in the United States

School buildings have a great potential for carbon emission reduction since their annual emission is about 72 million metric tons. Currently, more than 30% of school buildings were built before 1960 and are underperforming. To effectively reduce carbon emissions via school building retrofits, it is critical for policymakers to understand the carbon intensity reduction potential of retrofitting school buildings in different regions. Hence, this study develops a method to comprehensively assess the long-term carbon intensity reduction potential of aggregated commercial buildings on a county-by-county basis in the continental U.S. We apply this method to the K-12 school buildings including primary and secondary school buildings. Here, this paper predicts the carbon intensity reduction potential of K-12 school buildings with eight building retrofit measures from 2022 to 2050 in the continental U.S. The results reveal several interesting findings: (1) In the approximately 3,000 counties of the U.S. from 2022 to 2050, the carbon intensity reduction potential of retrofitting K-12 school buildings in each county ranges from 0.41 kg/m 2 to 40.00 kg/m 2 . (2) Even in the same climate zone, the trends of carbon intensity reduction potential from 2022 to 2050 are different depending on their electricity sources. For example, in a hot and humid climate zone, the carbon intensity reduction potential in Florida will decrease from 2044 to 2048. However, in Mississippi, the carbon intensity reduction potential from 2044 to 2046 will increase due to the termination of the nuclear energy usage. (3) Generally, reducing lighting power density leads to more carbon intensity reduction in most states, but it might not be applicable for states with high clean energy penetration, such as Washington.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Predictive model using artificial neural network to design phase change material-based ocean thermal energy harvesting systems for powering uncrewed underwater vehicles

Uncrewed Underwater Vehicles (UUVs) are a major beneficiary of the phase change material (PCM)-based ocean thermal energy harvesting technology for their mission needs. However, this technology relies on different parameters and energy conversion steps that could be critical to the general energy generation efficiency. Sea trials showed that the design performed lower than their laboratory design specifications. This underperformance results from different factors, mainly the UUV’s trajectory, travel time, underwater ocean currents, temperature fluctuations, and biofouling on the heat exchanger due to long term underwater operations. Therefore, there exists a need to continuously monitor the ambient energy harvesting system and predict system performance, for mission planning purposes. Two major parameters influencing the energy harvesting system include the final pressure inside the hydraulic energy storage vessel or accumulator, and the electrical load value. Here, this work focuses on the hydraulic to electric energy conversion system. Therefore, a combination of numerical model and experimental testing is used to develop a predictive model using artificial neural network using MATLAB. After validation with experimental testing, 1000 data samples obtained from the numerical model are used to train the ANN. Compared to the experimental results, the developed ANN model can predict in less than a second the designed benchtop system’s total efficiency with less than 15 percent maximum error range. This predictive model development represents a cost-effective way for optimization and a computational energy efficient mode aboard UUVs for mission planning for deployed UUVs using PCM-based ocean thermal energy harvesting technology.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2020

Hydraulic fracturing to enhance injectivity and storage capacity of CO 2 storage reservoirs: Benefits and risks

Several potential CO 2 storage reservoirs have been found to have insufficient porosity and permeability to support cost effective commercial-scale injection. As a result, the use of hydraulic fracturing to enhance injectivity and storage capacity of CO 2 storage reservoirs was explored. Previous modeling studies indicate that fracturing can increase storage capacity by modest to significant amounts (10%–35%), depending on model assumptions. Simulations completed as part of this study confirm that for a range of horizontal well lengths, number of fractures, fracture geometries, and fracture properties, injectivity is improved and capacity increases by 13%–71% over the base case unfractured vertical well. Intuitively, increasing the well length and number of fractures had a corresponding impact on increased capacity. Here, fracture area (i.e., the fracture height multiplied by its width) was an important parameter for increasing capacity but the specific geometry (e.g., the ratio of height-to-width) was unimportant. The most important aspect that affected capacity was the ability of a fracture to connect high permeability horizontal zones in the reservoir. This would allow a single well to access both layers, thereby maximizing total storage capacity of the reservoir while likely leading to an overall increase of the CO 2 footprint, which is an important consideration for Class VI UIC permitting. The results of this work demonstrate that hydraulic fracturing is an attractive option to consider when faced with an underperforming geologic carbon storage site that is at risk of causing a project to fail.

03 NATURAL GAS↗
DOE OSTI2023

Semihard magnetic properties of TiFe 2.5 iron-rich Laves phase and the effect of 4d- and 5d-element-substitutions for Ti

Rare-earth-free compounds exhibiting modest intrinsic hard magnetic properties may still yield viable permanent magnets if their constituent elements are abundant and inexpensive, and the properties are at least comparable to those of the hard ferrites. In this study, one such compound, the off-stoichiometric TiFe 2.5 Laves phase with the hexagonal C14 crystal structure, was found to exhibit – in addition to the already known room-temperature ferromagnetism – a uniaxial, albeit weak, magnetic anisotropy. With a Curie temperature of 422 K, saturation magnetization of 55.3 Am 2 /kg and magnetic hardness parameter of 0.97 this semihard compound is just below the threshold for being of interest for the development into permanent magnets. Replacing a small fraction of Ti with the 4d Nb and Mo or with the 5d Ta or W increases the anisotropy field, but only below room temperature. Replacing Ti with Zr or Hf increases the Curie temperature and leads to a spin reorientation below 200 K. All these substitutions, as well as combined (Zr,W) and (Nb,W) substitutions, fail to improve the room-temperature intrinsic hard magnetic properties of the Fe-rich Laves phase. Furthermore, an attempt to develop a room-temperature coercivity through high-energy ball-milling yielded a value of 0.026 T, an unusually small 2.9% fraction of the anisotropy field. Here, defects inherent in the C14 crystal lattice may be responsible for the underperformance.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Improved thermoelectric performance in Cr-doped two-dimensional Bi 2 Te 3

Thermoelectric materials with high electrical conductivity and low thermal conductivity (e.g., Bi 2 Te 3 ) can efficiently convert waste heat into electricity. However, despite favorable theoretical predictions, individual Bi 2 Te 3 nanostructures such as two-dimensional (2D) nanoplates tend to underperform bulk Bi 2 Te 3 . We report a novel surface doping technique to synthesize highly n-type Bi 2 Te 3 nanoplates using an external Cr coating followed by a thermal annealing process in a reducing atmosphere, as well as the mechanism by which this surface coating – only a few atoms or less in thickness – can observably impact the thermoelectric performance of 2D Bi 2 Te 3 . The Cr atoms act as n-type carrier donors by directly incorporating into the Bi 2 Te 3 structure during thermal annealing, enhancing electrical conductivity by ∼ 70 % while increasing thermal conductivity by only ∼ 5 % at room temperature. Compared to the uncoated Bi 2 Te 3 nanoplate, the Cr-doped Bi 2 Te 3 nanoplate exhibits a doubled thermoelectric figure of merit (zT), which is still relatively low. Raman spectroscopy and chemical potential simulations further confirm that Cr atoms are incorporated into the Bi 2 Te 3 structure.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Comprehensive assessment of metrology techniques for heliostat efficiency and performance evaluation

Concentrating solar power plants, specifically central receiver type systems and their heliostat field, are struggling with negative reputation in the USA, due to perceived underperformance and reliability issues. This is in part due to a lack of standards for performance assessment as well as overly simplified techno-economical models. A better understanding of influences and losses along the solar radiation path from the sun, across the solar collector to the receiver, increases the fidelity of heliostat efficiency assessment as well as solar field performance predictions. Such data are currently scarce and require a complete set of metrology capabilities to evaluate direct solar irradiance, sun shape, atmospheric attenuation, reflectance, collector shape, slope errors and total beam dispersion. In preparation for establishing a 3rd party metrology platform in collaboration with Sandia National Labs, NLR conducted a scoping study on available metrology. We present an extensive overview of techniques and commercial systems for each category. Our work includes an analysis to increase understanding of strengths and limitations of the many techniques used for surface shape and slope measurement. This applies to a controlled, indoor or outdoor laboratory environment assessing a single heliostat.

14 SOLAR ENERGY↗
DOE OSTI2026

Hierarchical Reinforcement Learning of a Short-Range Bond-Order Potential for Silica: Analytic Embedding of Coordination with Classical Efficiency

Reinforcement learning (RL) has recently emerged as a data-efficient strategy to parametrize short-range interatomic potentials. Building on our past RL optimization of pairwise silica models, we extend the framework to a bond-order (Tersoff-type) potential that provides an analytic embedding of local coordination through a three-body term. A hierarchical RL workflow combining continuous-action Monte Carlo Tree Search and property-based rewards efficiently explores the 26-dimensional parameter space, sequentially optimizing lattice parameters, densities, angles, and cohesive energies of 21 silica polymorphs. The resulting models, Q-Tersoff and ML-Tersoff, reproduce the energetic ordering of low-energy phases and capture the angular correlations and amorphous structure factors of silica with improved fidelity over pairwise force fields, while remaining orders of magnitude faster than high-dimensional machine-learned potentials. Both models underperform for elastic constants and high-energy frameworks, delineating the limits of the current analytic form. The approach establishes a general and interpretable route to angle-aware, short-range potentials that bridge physics-based and machine-learned descriptions of silicate materials.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Influence of Linker Identity on the Photochemistry of Uranyl-Organic Frameworks

While uranyl-based metal–organic frameworks (MOFs) boast impressive photocatalytic abilities, significant questions remain regarding their excitation pathways and methods to fine-tune their performance due to the lack of information regarding heterogeneous uranyl catalysis. Herein, we investigated how linker identity and photoexcitation impact uranyl photocatalysis when the uranyl coordination environment remains constant. Toward this end, we prepared three uranyl-based MOFs (NU-1301, NU-1307, and ZnTCPP-U2) and then examined the structural and photochemical properties of each through X-ray diffraction, X-ray absorption, and photoluminescence. We then correlated our observations to the photocatalytic performance for fluorination of cyclooctane. The excitation profile from NU-1301 and NU-1307 exhibited spin-forbidden linker transitions and uranyl vibronic progressions, with uranyl excitation and emission being most dominant in NU-1301. Consequently, NU-1301 was a more active photocatalyst than NU-1307. In contrast, the excitation profile from ZnTCPP-U2 contained transitions associated with the porphyrin linker exclusively. Photocatalytic activity from ZnTCPP-U2 significantly underperformed in comparison to that of the other two MOFs. Furthermore, these data suggest that linkers’ photophysical properties can be used to predict the photocatalytic behavior of uranyl-containing MOFs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Enabling High Stability of Co-Free LiNiO 2 Cathode via a Sulfide-Enriched Cathode Electrolyte Interface

Cobalt-free lithium nickel oxide (LNO) has garnered significant interest as the end member of high-nickel layered oxide cathodes for next-generation batteries. However, its practical performance notably underperforms expectations because of the structural degradation and unstable interfacial chemistry with electrolytes during cycling. Here, in this study, we report that a durable cathode-electrolyte interface (CEI), enriched by in situ formed sulfides and borides, can inhibit LNO structural degradation and suppress Ni ion dissolution. With the CEI protection, the stability of LNO can be remarkably extended, and batteries demonstrate a capacity retention rate of 84% (30 °C) and 79% (50 °C) after 200 cycles at 1C, respectively. These results demonstrate that enriching CEI with sulfur-containing species can effectively stabilize the interfacial chemistry of LNO, particularly at an elevated temperature of 50 °C. This finding provides valuable perspectives on designing electrolytes for cobalt-free LNO and other high-Ni cathodes toward the development of next-generation high-energy-density lithium-ion batteries.

25 ENERGY STORAGE↗
DOE OSTI2022

Heteroanionic Control of Exemplary Second-Harmonic Generation and Phase Matchability in 1D LiAsS 2– x Se x

The isostructural heteroanionic compounds β-LiAsS 2–x Se x (x = 0, 0.25, 1, 1.75, 2) show a positive correlation between selenium content and second-harmonic response and greatly outperform the industry standard AgGaSe 2 . These materials crystallize in the noncentrosymmetric space group Cc as one-dimensional $\frac{1}{∞}$ [AsQ 2 ] - (Q = S, Se, S/Se) chains consisting of corner-sharing AsQ3 trigonal pyramids with charge-balancing Li + atoms interspersed between the chains. LiAsS 2–x Se x melts congruently for 0 ≤ x ≤ 1.75, but when the Se content exceeds x = 1.75, crystallization is complicated by a phase transition. This behavior is attributed to the β- to α-phase transition present in LiAsSe 2 , which is observed in the Se-rich compositions. The band gap decreases with increasing Se content, starting at 1.63 eV (LiAsS 2 ) and reaching 1.06 eV (β-LiAsSe 2 ). Second-harmonic generation measurements as a function of wavelength on powder samples of β-LiAsS 2–x Se x show that these materials exhibit significantly higher nonlinearity than AgGaSe 2 (d 36 = 33 pm/V), reaching a maximum of 61.2 pm/V for LiAsS 2 . In comparison, single-crystal measurements for LiAsSSe yielded a d eff = 410 pm/V. LiAsSSe, LiAsS 0.25 Se 1.75 , and β-LiAsSe 2 show phase-matching behavior for incident wavelengths exceeding 3 μm. The laser-induced damage thresholds from two-photon absorption processes are on the same order of magnitude as AgGaSe 2 , with S-rich materials slightly outperforming AgGaSe 2 and Se-rich materials slightly underperforming AgGaSe 2 .

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Atmospheric Structure Prediction for Infrasound Propagation Modeling Using Deep Learning

Abstract Infrasound is generated by a variety of natural and anthropogenic sources. Infrasonic waves travel through the dynamic atmosphere, which can change on the order of minutes to hours. Infrasound propagation largely depends on the wind and temperature structure of the atmosphere. Numerical weather prediction models are available to provide atmospheric specifications, but uncertainties in these models exist and they are computationally expensive to run. Machine learning has proven useful in predicting tropospheric weather using Long Short‐Term Memory (LSTM) networks. An LSTM network is utilized to make atmospheric specification predictions up to ∼30 km for three different training and testing scenarios: (a) the model is trained and tested using only radiosonde data from the Albuquerque, NM, USA station, (b) the model is trained on radiosonde stations across the contiguous US, excluding the Albuquerque, NM, USA station, which was reserved for testing, and (c) the model is trained and tested on radiosonde stations across the contiguous US. Long Short‐Term Memory predictions are compared to a state‐of‐the‐art reanalysis model and show cases where the LSTM outperforms, performs equally as well, or underperforms in comparison to the state‐of‐the‐art. Regional and temporal trends in model performance across the US are also discussed. Results suggest that the LSTM model is a viable tool for predicting atmospheric specifications for infrasound propagation modeling.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

A Mass Conservation Relaxed (MCR) LSTM Model for Streamflow Simulation Across CONUS

The recent development of the physics-aware Mass-Conserving Long Short-Term Memory network (MC-LSTM) provides an alternative to other data-driven Deep Learning (DL) models in hydrology. Mass-Conserving Long Short-Term Memory incorporates mass conservation directly into the LSTM architecture. Despite the theoretical advancements, studies have reported a surprisingly limited performance of the MC-LSTM in streamflow simulation. We hypothesize that such a limitation is due to the unrealistic mass conservation scheme in MC-LSTM, which overlooks unobserved incoming water fluxes beyond precipitation. As an attempt to verify this hypothesis, we propose a Mass Conservation Relaxed LSTM (MCR-LSTM), which incorporates a bi-directional mass relaxation (MR) component to account for potential incoming water fluxes beyond precipitation. We train and test the proposed MCR-LSTM model across 531 watersheds in the contiguous United States (CONUS) against three baseline models: the Sacramento Soil Moisture Accounting, LSTM, and MC-LSTM. Our results show that MCR-LSTM outperforms MC-LSTM despite its underperformance compared to LSTM. Specifically, MCR-LSTM's advantage over MC-LSTM is mainly seen in the Plains and Western U.S., where the newly incorporated MR component better simulates water loss and suggests the likely existence of additional incoming water fluxes beyond precipitation, respectively. The novelty and contribution of this study are twofold: firstly, it introduces an alternative physics-aware DL tool (i.e., MCR-LSTM) in hydrology with higher accuracy in specific regions compared to MC-LSTM. Secondly, it provides a diagnosis of regions where strict, precipitation-based mass conservation constraints may be unrealistic in streamflow simulation.

deep learning↗