Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “model structural error”

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 883 records · Page 49
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Development and experimental evaluation of new building air leakage measurement methods: measurement of interior air leaks and comparison to conventional methods

Building air leaks (both through exterior and interior surfaces) can have a significant impact on energy consumption, indoor air quality, fire safety, and moisture accumulation affecting structural durability. Blower door testing has been used to measure leaks in buildings, but commonly used testing methods do not directly measure interior leaks. In this paper, new testing methods (guarded interior test and zonal multipoint pressure testing method) are presented that directly measure these interior leaks, utilizing common blower door equipment for both single and multi-point testing. Furthermore, these new methods are compared to conventional methods in terms of the information provided, limitations and time/effort needed. In addition, building leak measurement results are analyzed to reveal a) coupling between power law model values (exponent and coefficient) for an ensemble of buildings, b) the error in using single point testing when estimating low pressure leakage, and c) how building power law models vary from low to high pressure ranges.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2016

TREAT neutronics analysis and design support, Part II: Multi-SERTTA-CAL

Experiment vehicle design is necessary in preparation for Transient Reactor Test (TREAT) facility restart and the resumption of transient testing to support Accident Tolerant Fuel (ATF) characterization and other future fuels testing requirements. Currently the most mature vehicle design is the Multi-SERTTA (Static Environments Rodlet Transient Test Apparatuses), which can accommodate up to four concurrent rodlet-sized specimens under separate environmental conditions. Robust test vehicle design requires neutronics analyses to support design development, optimization of the power coupling factor (PCF) to efficiently maximize energy generation in the test fuel rodlets, and experiment safety analyses. An integral aspect of prior TREAT transient testing was the incorporation of calibration experiments to experimentally evaluate and validate test conditions in preparation of the actual fuel testing. The calibration experiment package established the test parameter conditions to support fine-tuning of the computational models to deliver the required energy deposition to the fuel samples. The calibration vehicle was designed to be as near neutronically equivalent to the experiment vehicle as possible to minimize errors between the calibration and final tests. The Multi-SERTTA-CAL vehicle was designed to serve as the calibration vehicle supporting Multi-SERTTA experimentation. Models of the Multi-SERTTA-CAL vehicle containing typical PWR-fuel rodlets were prepared and neutronics calculations were performed using MCNP6.1 with ENDF/B-VII.1 nuclear data libraries; these results were then compared against those performed for Multi-SERTTA to determine the similarity and possible design modification necessary prior to construction of these experiment vehicles. The estimated reactivity insertion worth into the TREAT core is very similar between the two vehicle designs, with the primary physical difference being a hollow Inconel tube running down the length of the calibration vehicle. Calculations of PCF indicate that on average there is a reduction of approximately 6 % for PWR fuel rodlets irradiated under both wet and dry conditions. Changes to the primary or secondary vessel structure in the calibration vehicle can be performed to offset this discrepancy and maintain neutronic equivalency. Current possible modifications to the calibration vehicle include reduction of the primary vessel wall thickness, swapping Zircaloy-4 for stainless steel 316 in the secondary containment, or slight modification to the temperature and pressure of the water environment within the primary vessel. Removal of some of the instrumentation within the calibration vehicle can also serve to slightly increase the PCF. Future efforts include further modification and optimization of the Multi-SERTTA and Multi-SERTTA-CAL designs in preparation of actual TREAT transient testing. Experimental results from both test vehicles will be compared against calculational results and methods to provide validation and support additional neutronics analyses. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2024

Boosting Barlow Twins Reduced Order Modeling for Machine Learning‐Based Surrogate Models in Multiphase Flow Problems

Abstract We present an innovative approach called boosting Barlow Twins reduced order modeling (BBT‐ROM) to enhance the reliability of machine learning surrogate models for multiphase flow problems. BBT‐ROM builds upon Barlow Twins reduced order modeling that leverages self‐supervised learning to effectively handle linear and nonlinear manifolds by constructing well‐structured latent spaces of input parameters and output quantities. To address the challenge of high contrast data in multiphase flow problems due to injection wells and faults, we employ a boosting algorithm within BBT‐ROM. This algorithm sequentially trains a set of weak models (i.e., inaccurate models), improving prediction accuracy through ensemble learning. To evaluate the performance of BBT‐ROM, we conduct three three‐dimensional multiphase flow problems, including waterflooding and geologic carbon storage (GCS), with varying numbers of input parameter cases and model domain features. The results demonstrate that BBT‐ROM excels at predicting non‐wetting phase saturation (e.g., oil or saturation) and fluid pressure, with average relative errors ranging from 0.5% to 3%. Importantly, BBT‐ROM showcases robustness when faced with limited input parameter space during GCS testing.

58 GEOSCIENCES↗
DOE OSTI2024

RhizoNet segments plant roots to assess biomass and growth for enabling self-driving labs

Abstract Flatbed scanners are commonly used for root analysis, but typical manual segmentation methods are time-consuming and prone to errors, especially in large-scale, multi-plant studies. Furthermore, the complex nature of root structures combined with noisy backgrounds in images complicates automated analysis. Addressing these challenges, this article introduces RhizoNet, a deep learning-based workflow to semantically segment plant root scans. Utilizing a sophisticated Residual U-Net architecture, RhizoNet enhances prediction accuracy and employs a convex hull operation for delineation of the primary root component. Its main objective is to accurately segment root biomass and monitor its growth over time. RhizoNet processes color scans of plants grown in a hydroponic system known as EcoFAB, subjected to specific nutritional treatments. The root detection model using RhizoNet demonstrates strong generalization in the validation tests of all experiments despite variable treatments. The main contributions are the standardization of root segmentation and phenotyping, systematic and accelerated analysis of thousands of images, significantly aiding in the precise assessment of root growth dynamics under varying plant conditions, and offering a path toward self-driving labs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Mechanism of structural colors in binary mixtures of nanoparticle-based supraballs

Inspired by structural colors in avian species, various synthetic strategies have been developed to produce noniridescent, saturated colors using nanoparticle assemblies. Nanoparticle mixtures varying in particle chemistry and size have additional emergent properties that affect the color produced. For complex multicomponent systems, understanding the assembled structure and a robust optical modeling tool can empower scientists to identify structure-color relationships and fabricate designer materials with tailored color. Here, we demonstrate how we can reconstruct the assembled structure from small-angle scattering measurements using the computational reverse-engineering analysis for scattering experiments method and use the reconstructed structure in finite-difference time-domain calculations to predict color. We successfully, quantitatively predict experimentally observed color in mixtures containing strongly absorbing nanoparticles and demonstrate the influence of a single layer of segregated nanoparticles on color produced. The versatile computational approach that we present is useful for engineering synthetic materials with desired colors without laborious trial-and-error experiments.

36 MATERIALS SCIENCE↗
DOE OSTI2013

Structural health and prognostics management for offshore wind turbines :

Operations and maintenance costs for offshore wind plants are significantly higher than the current costs for land-based (onshore) wind plants. One way to reduce these costs would be to implement a structural health and prognostic management (SHPM) system as part of a condition based maintenance paradigm with smart load management and utilize a state-based cost model to assess the economics associated with use of the SHPM system. To facilitate the development of such a system a multi-scale modeling approach developed in prior work is used to identify how the underlying physics of the system are affected by the presence of damage and faults, and how these changes manifest themselves in the operational response of a full turbine. This methodology was used to investigate two case studies: (1) the effects of rotor imbalance due to pitch error (aerodynamic imbalance) and mass imbalance and (2) disbond of the shear web; both on a 5-MW offshore wind turbine in the present report. Based on simulations of damage in the turbine model, the operational measurements that demonstrated the highest sensitivity to the damage/faults were the blade tip accelerations and local pitching moments for both imbalance and shear web disbond. The initial cost model provided a great deal of insight into the estimated savings in operations and maintenance costs due to the implementation of an effective SHPM system. The integration of the health monitoring information and O&M cost versus damage/fault severity information provides the initial steps to identify processes to reduce operations and maintenance costs for an offshore wind farm while increasing turbine availability, revenue, and overall profit.

Myrent, Noah J.↗
DOE OSTI2016

Verification test of the SURF and SURFplus models in xRage

As a verification test of the SURF and SURFplus models in the xRage code we use a propagating underdriven detonation wave in 1-D. This is about the only test cases for which an accurate solution can be determined based on the theoretical structure of the solution. The solution consists of a steady ZND reaction zone profile joined with a scale invariant rarefaction or Taylor wave and followed by a constant state. The end of the reaction profile and the head of the rarefaction coincide with the sonic CJ state of the detonation wave. The constant state is required to match a rigid wall boundary condition. For a test case, we use PBX 9502 with the same EOS and burn rate as previously used to test the shock detector algorithm utilized by the SURF model. The detonation wave is propagated for 10 μs (slightly under 80mm). As expected, the pointwise errors are largest in the neighborhood of discontinuities; pressure discontinuity at the lead shock front and pressure derivative discontinuities at the head and tail of the rarefaction. As a quantitative measure of the overall accuracy, the L2 norm of the difference of the numerical pressure and the exact solution is used. Results are presented for simulations using both a uniform grid and an adaptive grid that refines the reaction zone.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗
DOE OSTI2020

Nano-engineering the material structure of preferentially oriented nano-graphitic carbon for making high-performance electrochemical micro-sensors

Direct synthesis of thin-film carbon nanomaterials on oxide-coated silicon substrates provides a viable pathway for building a dense array of miniaturized (micron-scale) electrochemical sensors with high performance. However, material synthesis generally involves many parameters, making material engineering based on trial and error highly inefficient. Here, we report a two-pronged strategy for producing engineered thin-film carbon nanomaterials that have a nano-graphitic structure. First, we introduce a variant of the metal-induced graphitization technique that generates micron-scale islands of nano-graphitic carbon materials directly on oxide-coated silicon substrates. A novel feature of our material synthesis is that, through substrate engineering, the orientation of graphitic planes within the film aligns preferentially with the silicon substrate. This feature allows us to use the Raman spectroscopy for quantifying structural properties of the sensor surface, where the electrochemical processes occur. Second, we find phenomenological models for predicting the amplitudes of the redox current and the sensor capacitance from the material structure, quantified by Raman. Our results indicate that the key to achieving high-performance micro-sensors from nano-graphitic carbon is to increase both the density of point defects and the size of the graphitic crystallites. Our study offers a viable strategy for building planar electrochemical micro-sensors with high-performance.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2019

Approximate Latent Factor Algorithm for Scenario Selection and Weighting in Transmission Expansion Planning

One major difficulty in transmission expansion planning is selecting the representative scenarios to use to evaluate candidate transmission networks. The variability in demand and renewable generation makes the inclusion of several scenarios critical when calculating reliability and cost, but including too many scenarios in an optimization is computationally intractable. To reduce the number of representative operating conditions needed to obtain an accurate approximation of the system, we propose a method rooted in multivariate statistics that exploits the latent correlative structure between different scenarios and network configurations. The proposed algorithm includes an objective and rigorous way to select a subset of scenarios that provide as much information about the system as possible, and a method to accurately approximate the system cost from that scenario subset. The result is a set of scenarios and weights that are easily incorporated into traditional transmission expansion planning formulations. We apply this to a 312-bus WECC model with 8,736 distinct operating conditions. The transmission plans found with the proposed method are more reliable and have a lower total cost than those from other scenario reduction techniques, as well as a smaller error between the expected system performance from the optimization objective and the actual system performance.

42 ENGINEERING↗
DOE OSTI2024

Methyl α‐D‐galactopyranosyl‐(1→3)‐β‐D‐galactopyranoside and methyl β‐D‐galactopyranosyl‐(1→3)‐β‐D‐galactopyranoside: Glycosidic linkage conformation determined from MA'AT analysis

Abstract MA'ATanalysis has been applied to two biologically‐importantO‐glycosidic linkages in two disaccharides, α‐D‐Galp‐(1→3)‐β‐D‐GalpOMe (3) and β‐D‐Galp‐(1→3)‐β‐D‐GalpOMe (4). Using density functional theory (DFT) to obtain parameterized equations relating a group of trans‐O‐glycosidic NMR spin‐couplings to eitherphi(ϕ') orpsi(ψ'), and experimental 3 J COCH , 2 J COC , and 3 J COCC spin‐couplings measured in aqueous solution in 13 C‐labeled isotopomers, probability distributions ofϕ'andψ'in each linkage were determined and compared to those determined by aqueous 1‐μs molecular dynamics (MD) simulation. Good agreement was found between theMA'ATand single‐state MD conformational models of these linkages for the most part, with modest (approximately <15°) differences in the mean values ofϕ'andψ', although the envelope of allowed angles (encoded in circular standard deviations or CSDs) is consistently larger forϕ'determined fromMA'ATanalysis than from MD for both linkages. TheMA'ATmodel of the α‐Galp‐(1→3)‐β‐Galplinkage agrees well with those determined previously using conventional NMR methods ( 3 J COCH values and/or 1 H‐ 1 H NOEs), but some discrepancy was observed for the β‐Galp‐(1→3)‐β‐Galplinkage, which may arise from errors in the conventions used to describe the linkage torsion angles. Statistical analyses of X‐ray crystal structures show ranges ofϕ'andψ'for both linkages that include the mean angles determined fromMA'ATanalyses, although both angles adopt a wide range of values in the crystalline state, withϕ'in β‐Galp‐(1→3)‐β‐Galplinkages showing greater‐than‐expected conformational variability.

Chemistry↗
DOE OSTI2024

Improving precision and accuracy of genetic mapping with genotyping‐by‐sequencing data in outcrossing species

Abstract Genotyping‐by‐sequencing (GBS) is a widely used strategy for obtaining large numbers of genetic markers in model and non‐model organisms. In crop plants, GBS‐derived marker datasets are frequently used to perform quantitative trait locus (QTL) mapping. In some plant species, however, high heterozygosity and complex genome structure mean that researchers must use care in handling GBS data to conduct QTL mapping most effectively. Such outbred crops include most of the perennial grass and tree species used for bioenergy. To identify strategies for increasing accuracy and precision of QTL mapping using GBS data in outbred crops, we conducted an empirical study of SNP‐calling and genetic map‐building pipeline parameters in a Miscanthus sinensis population, and a complementary simulation study to estimate the relationship between genome‐wide error rate, read depth, and marker number. The bioenergy grass Miscanthus is an obligate outcrossing species with a recent (diploidized) whole‐genome duplication. For the study of empirical M. sinensis data, we compared two SNP‐calling methods (one non‐reference‐based and one reference‐based), a series of depth filters (12×, 20×, 30×, and 40×) and two map‐construction methods (i.e., marker ordering: linkage‐only and order‐corrected based on a reference genome). We found that correcting the order of markers on a linkage map by using a high‐quality reference genome improved QTL precision (shorter confidence intervals). For typical GBS datasets of between 1000 and 5000 markers to build a genetic map for biparental populations, a depth filter set at 30× to 40× applied to outbred populations provided a genome‐wide genotype‐calling error rate of less than 1%, improved accuracy of QTL point estimates and minimized type I errors for identifying QTL. Based on these results, we recommend using a reference genome to correct the marker order of genetic maps and a robust genotype depth filter to improve QTL mapping for outbred crops.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2014

Structural Health and Prognostics Management for Offshore Wind Turbines: Sensitivity Analysis of Rotor Fault and Blade Damage with O&M Cost Modeling

Operations and maintenance costs for offshore wind plants are significantly higher than the current costs for land-based (onshore) wind plants. One way to reduce these costs would be to implement a structural health and prognostic management (SHPM) system as part of a condition based maintenance paradigm with smart load management and utilize a state-based cost model to assess the economics associated with use of the SHPM system. To facilitate the development of such a system a multi-scale modeling and simulation approach developed in prior work is used to identify how the underlying physics of the system are affected by the presence of damage and faults, and how these changes manifest themselves in the operational response of a full turbine. This methodology was used to investigate two case studies: (1) the effects of rotor imbalance due to pitch error (aerodynamic imbalance) and mass imbalance and (2) disbond of the shear web; both on a 5-MW offshore wind turbine in the present report. Sensitivity analyses were carried out for the detection strategies of rotor imbalance and shear web disbond developed in prior work by evaluating the robustness of key measurement parameters in the presence of varying wind speeds, horizontal shear, and turbulence. Detection strategies were refined for these fault mechanisms and probabilities of detection were calculated. For all three fault mechanisms, the probability of detection was 96% or higher for the optimized wind speed ranges of the laminar, 30% horizontal shear, and 60% horizontal shear wind profiles. The revised cost model provided insight into the estimated savings in operations and maintenance costs as they relate to the characteristics of the SHPM system. The integration of the health monitoring information and O&M cost versus damage/fault severity information provides the initial steps to identify processes to reduce operations and maintenance costs for an offshore wind farm while increasing turbine availability, revenue, and overall profit.

17 WIND ENERGY↗
DOE OSTI2025

Scaled-up fabrication of durable and porous adsorbent-coated minichannels on aluminum for CO 2 separation

A method was developed to fabricate zeolite 13X adsorbent-coated minichannels on aluminum for CO₂ adsorption applications in this study. It emphasizes the innovative use of aluminum as a substrate, which offers airtight assembly, paving the way for highly efficient adsorption systems. An optimized coating process was developed using a slurry of Zeolite 13X, yeast, sugar, and xanthan gum, resulting in durable and highly porous layers that enhance CO₂ capture performance. A PETG peeler was designed to remove the top layer of the yeast-engineered adsorbent coatings, revealing a super porous and foamy structure. The teeth of the peeler were designed and fabricated for high repeatability and rapid prototyping. Breakthrough experiments were conducted on the scaled-up adsorbent bed using gas mixtures of 80% CO₂, 20% N₂, and 20% CO₂, 80% N₂ to represent different industrial scenarios. The performance of the bed was evaluated at flow rates of 160 and 190 cm³ min -1 using a Raman Laser Gas Analyzer (RLGA), demonstrating stable adsorption without degradation across multiple cycles. Computational modeling of integral transport phenomena under the chosen experimental conditions was pursued using gPROMS ProcessBuilder™, and the modeling results were compared with those from the tests for adsorption time, which resulted in an error margin of 2% to 9% for the breakthrough time, confirming the easy reproducibility of the design through modeling. This research advances CO₂ capture technologies by providing an effective and scalable solution for producing aluminum-based adsorbent coated beds, supporting industrial carbon capture efforts.

CO2 capture↗
DOE OSTI2014

Cluster of Asphaltene Nanoaggregates by DC Conductivity and Centrifugation

A model of the dominant molecular and stable colloidal structures of asphaltenes has been proposed, the Yen–Mullins model. The formation of clusters of asphaltene nanoaggregates in toluene was reported elsewhere to occur at a concentration of a few grams per liter with a cluster aggregation number of approximately 8. In this work, we measure the DC-conductivity signal of toluene as a function of asphaltene concentration obtaining support for the critical clustering concentration (CCC) of a roughly 1.7 g/L in toluene. In addition, the small change in the Stokes drag at the CCC indicates that the cluster aggregation number is small, less than 10. The temperature variation of the CCC is measured to be small and within error, suggesting that cluster formation is entropically driven. Centrifugation experiments were also performed on asphaltene–toluene solutions at different concentrations. These experiments confirmed that a significant change in asphaltene aggregation occurs in the concentration range of the CCC. The CCC values from centrifugation and DC-conductivity measurements are roughly the same. The centrifugation experiments confirm a cluster size of ~5 nm corroborating the small aggregation number found in the DC-conductivity experiments. These results add to the growing body of literature validating the Yen–Mullins model.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2015

Human HLTF mediates postreplication repair by its HIRAN domain-dependent replication fork remodelling

Defects in the ability to respond properly to an unrepaired DNA lesion blocking replication promote genomic instability and cancer. Human HLTF, implicated in error-free replication of damaged DNA and tumour suppression, exhibits a HIRAN domain, a RING domain, and a SWI/SNF domain facilitating DNA-binding, PCNA-polyubiquitin-ligase, and dsDNA-translocase activities, respectively. Here, we investigate the mechanism of HLTF action with emphasis on its HIRAN domain. We found that in cells HLTF promotes the filling-in of gaps left opposite damaged DNA during replication, and this postreplication repair function depends on its HIRAN domain. Our biochemical assays show that HIRAN domain mutant HLTF proteins retain their ubiquitin ligase, ATPase and dsDNA translocase activities but are impaired in binding to a model replication fork. These data and our structural study indicate that the HIRAN domain recruits HLTF to a stalled replication fork, and it also provides the direction for the movement of the dsDNA translocase motor domain for fork reversal. We suggest functional similarities between the HIRAN, the OB, the HARP2, and other domains found in certain motor proteins, which may explain why only a subset of DNA translocases can carry out fork reversal.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2018

CRADA Final Report: Weld Predictor App

Welding is an important manufacturing process used in a broad range of industries and market sectors, including automotive, aerospace, heavy manufacturing, medical, and defense. During welded fabrication, high localized heat input and subsequent rapid cooling result in the creation of residual stresses and distortion. These residual stresses can significantly affect the fatigue resistance, cracking behavior, and load-carrying capacity of welded structures during service. Further, additional fitting and tacking time is often required to fit distorted subassemblies together, resulting in non-value added cost. Using trial-and-error methods to determine which welding parameters, welding sequences, and fixture designs will most effectively reduce distortion is a time-consuming and expensive process. For complex structures with many welds, this approach can take several months. For this reason, efficient and accurate methods of mitigating distortion are in-demand across all industries where welding is used. Analytical and computational methods and commercial software tools have been developed to predict welding-induced residual stresses and distortion. Welding process parameters, fixtures, and tooling can be optimized to reduce the HAZ softening and minimize weld residual stress and distortion, improving performance and reducing design, fabrication and testing costs. However, weld modeling technology tools are currently accessible only to engineers and designers with a background in finite element analysis (FEA) who work with large manufacturers, research institutes, and universities with access to high-performance computing (HPC) resources. Small and medium enterprises (SMEs) in the US do not typically have the human and computational resources needed to adopt and utilize weld modeling technology. To allow an engineer with no background in FEA and SMEs to gain access to this important design tool, EWI and the Ohio Supercomputer Center (OSC) developed the online weld application software tool “WeldPredictor” ( https://eweldpredictor.ewi.org ). About 1400 users have tested this application. This project marked the beginning of development on the next version of WeldPredictor that addresses many outstanding features of the original, including 3D models, allow more material hardening laws, model material phase transformation, and uses open source finite element solvers to quickly solve problems (as opposed to expensive commercial tools).

42 ENGINEERING↗
DOE OSTI2020

Reactor Power Level Estimation by Fusing Multi-Modal Sensor Measurements

Estimates of the power level of a nuclear reactor based on measurements from an independent monitoring sensor system can help in the compliance verification of its declared operations. We present a three-level fusion method to estimate the power level of a nuclear reactor using features derived from infrared, electromagnetic, and acoustic sensor measurements collected in proximity to the reactor. Based on a simplified analytical model of the secondary coolant system of the reactor, we identify partial regression functions of the power level in terms of the temperature difference between inlet and outlet of coolant pipes, and the activity levels of four fans and four pumps, which are estimated as features from the sensor measurements. The power level estimator employs a combination of aggregate and complementary fusion steps at three levels to incorporate the multi-modal features in a structure that reflects the secondary cooling system and its partial regression functions. Using the measurements from a test campaign at an operational reactor, we show that this estimator achieves 3.47% or lower root mean square error under 5-fold cross validation. More generally, these results illustrate a progressive reduction in estimation error as additional modalities are appropriately incorporated, and that the fuser outperforms single modality features and their sub-combinations.

Rao, Nageswara↗
DOE OSTI2024

When ancient numerical demons meet physics-informed machine learning: adjoint-based gradients for implicit differentiable modeling

Recent advances in differentiable modeling, a genre of physics-informed machine learning that trains neural networks (NNs) together with process-based equations, have shown promise in enhancing hydrological models' accuracy, interpretability, and knowledge-discovery potential. Current differentiable models are efficient for NN-based parameter regionalization, but the simple explicit numerical schemes paired with sequential calculations (operator splitting) can incur numerical errors whose impacts on models' representation power and learned parameters are not clear. Implicit schemes, however, cannot rely on automatic differentiation to calculate gradients due to potential issues of gradient vanishing and memory demand. Here we propose a “discretize-then-optimize” adjoint method to enable differentiable implicit numerical schemes for the first time for large-scale hydrological modeling. The adjoint model demonstrates comprehensively improved performance, with Kling–Gupta efficiency coefficients, peak-flow and low-flow metrics, and evapotranspiration that moderately surpass the already-competitive explicit model. Therefore, the previous sequential-calculation approach had a detrimental impact on the model's ability to represent hydrological dynamics. Furthermore, with a structural update that describes capillary rise, the adjoint model can better describe baseflow in arid regions and also produce low flows that outperform even pure machine learning methods such as long short-term memory networks. The adjoint model rectified some parameter distortions but did not alter spatial parameter distributions, demonstrating the robustness of regionalized parameterization. Despite higher computational expenses and modest improvements, the adjoint model's success removes the barrier for complex implicit schemes to enrich differentiable modeling in hydrology.

58 GEOSCIENCES↗