Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “generalizability”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Leveraging artificial intelligence and advanced food processing techniques for enhanced food safety, quality, and security: a comprehensive review

Artificial intelligence is emerging as a transformative force in addressing the multifaceted challenges of food safety, food quality, and food security. This review synthesizes advancements in AI-driven technologies, such as machine learning, deep learning, natural language processing, and computer vision, and their applications across the food supply chain, based on a comprehensive analysis of literature published from 1990 to 2024. AI enhances food safety through real-time contamination detection, predictive risk modeling, and compliance monitoring, reducing public health risks. It improves food quality by automating defect detection, optimizing shelf-life predictions, and ensuring consistency in taste, texture, and appearance. Furthermore, AI addresses food security by enabling resource-efficient agriculture, yield forecasting, and supply chain optimization to ensure the availability and accessibility of nutritious food resources. This review also highlights the integration of AI with advanced food processing techniques such as high-pressure processing, ultraviolet treatment, pulsed electric fields, cold plasma, and irradiation, which ensure microbial safety, extend shelf life, and enhance product quality. Additionally, the integration of AI with emerging technologies such as the Internet of Things, blockchain, and AI-powered sensors enables proactive risk management, predictive analytics, and automated quality control. By examining these innovations' potential to enhance transparency, efficiency, and decision-making within food systems, this review identifies current research gaps and proposes strategies to address barriers such as data limitations, model generalizability, and ethical concerns. These insights underscore the critical role of AI in advancing safer, higher-quality, and more secure food systems, guiding future research and fostering sustainable food systems that benefit public health and consumer trust.

AI↗
DOE OSTI2026

Causality-respecting adaptive refinement for PINNs: enabling precise interface evolution in phase field modeling

Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving physical systems described by partial differential equations (PDEs). However, their accuracy in dynamical systems, particularly those involving sharp moving boundaries with complex initial morphologies, remains a challenge. Here, this study introduces an approach combining residual-based adaptive refinement (RBAR) with causality-informed training to enhance the performance of PINNs in solving spatio-temporal PDEs. Our method employs a three-step iterative process: initial causality-based training, RBAR-guided domain refinement, and subsequent causality training on the refined mesh. Applied to the Allen-Cahn equation, a widely-used model in phase field simulations, our approach demonstrates significant improvements in solution accuracy and computational efficiency over traditional PINNs. Notably, we observe an ‘overshoot and relocate’ phenomenon in dynamic cases with complex morphologies, showcasing the method’s adaptive error correction capabilities. This synergistic interaction between RBAR and causality training enables accurate capture of interface evolution, even in challenging scenarios where traditional PINNs fail. Our framework not only resolves the limitations of uniform refinement strategies but also provides a generalizable methodology for solving a broad range of spatio-temporal PDEs. The enhanced performance of the RBAR–causality combined framework demonstrates its strong potential for advancing PINN-based modeling of physical systems characterized by complex, evolving interfaces.

Allen-Cahn equations↗
DOE OSTI2025

Analytical gradient-based optimization of CALPHAD model parameters

The calibration of CALPHAD (CALculation of PHAse Diagrams) models involves the solution of a very challenging high-dimensional multiobjective optimization problem. Traditional approaches to parameter fitting predominantly rely on gradient-free methods, which while robust, are computationally inefficient and often scale poorly with model complexity. In this work, we introduce and demonstrate a generalizable framework for analytic gradient-based optimization of the parameters of the CALPHAD model enabled by the recently formalized Jansson derivative technique. This method allows for efficient evaluation of gradients of thermodynamic properties at equilibrium with respect to model parameters, even in the presence of arbitrarily complex internal degrees of freedom. Leveraging these semi-analytic gradients, we employ the conjugate gradient (CG) method to optimize thermodynamic model parameters for four binary alloy systems: Cu-Mg, Fe-Ni, Cr-Ni, and Cr-Fe. Across all systems, CG achieves comparable or superior optimality relative to Bayesian ensemble Markov Chain Monte Carlo (MCMC) with improvements in computational efficiency ranging from one to three orders of magnitude. Furthermore, our results establish a new paradigm for CALPHAD assessments in which high fidelity data-rich model calibration becomes tractable using deterministic gradient-informed algorithms.

CALPHAD↗
DOE OSTI2025

Results of an interlaboratory study on the working curve in vat photopolymerization II: Towards a standardized method

The working curve measurement in photopolymer additive manufacturing is a ubiquitous measure of the cure depth of a printing resin as a function of radiant exposure of light. The fit parameters from this measurement (the depth of light penetration D p and the critical exposure E c ) are used to evaluate and report a resin’s printability, optimize processing parameters, and inform print and resin quality control. Despite its widespread use, the working curve lacks a standard measurement method. Here, following up on our paper “Results of an Interlaboratory Study on the Working Curve in Vat Photopolymerization” from last year, an interlaboratory study on the working curve was performed using calibrated, reproducible, bandpass filtered light sources. With these light sources, the variability between labs in measured working curves was dramatically reduced from the initial interlaboratory study. Aggregate data from this experiment produced reliable D p and E c measurements at 385 nm of 39.2 ± 3.7 µm and 12.3 ± 3.0 mJ cm −2 , respectively. At 405 nm the values of D p and E c are 69.3 ± 3.8 µm and 17.9 ± 2.3 mJ cm −2 , respectively. The results are agnostic to the thickness measurement tool utilized by participants, ensuring broad applicability across laboratories. We also tested the generalizability of the proposed method of using a filtered light source by filtering a commercial 405 nm light source and obtaining a working curve in agreement with the aggregate data from the interlaboratory study. This interlaboratory study provides a basis for a documentary standard for the working curve, so that the entire photopolymer additive manufacturing industry can share reproducible and interoperable working curve data.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Hydrologic investigations of radar-rainfall error propagation to rainfall-runoff model hydrographs

Rainfall is arguably the most important yet most variable input for rainfall-runoff hydrologic models. In this study, the authors search for the characteristics of radar-rainfall estimates that are most important for skillful streamflow predictions. They perform comprehensive hydrologic investigations of radar-rainfall characteristics, including spatiotemporal resolution, radar range visibility, statistical characterization of rainfall variability, all vis-a-vis basin characteristics such as size and river network topology. Since the true rainfall fields are unknown, the authors exploit a paradigm of using two independently constructed radar-rainfall products i.e., Multi-Radar Multi-Sensor and IFC-ZR used operationally by the Iowa Flood Center (IFC). Using the distributed hydrologic model called the Hillslope-Link Model for the domain of the state of Iowa, they evaluate streamflow prediction at 140 USGS gauge stations that monitor rivers in Iowa. Through spatial and temporal rainfall aggregation experiments, the authors show that the impact of spatial and temporal resolution of rainfall is significant typically for smaller basins while starts reducing significantly for basins larger than 1,000 km 2 . Other rainfall characteristics they explored do not reveal a strong signature in the relationship of rainfall differences between the two products and hydrograph errors. However, exploring the product similarities rather than differences reveals that the basin-wide rainfall volume has the most significant effect on streamflow prediction. The results from this study are generalizable for all rainfall observing systems.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Redox oxide@molten salt as a generalized catalyst design strategy for oxidative dehydrogenation of ethane via selective hydrogen combustion

Here, the current study demonstrates a redox oxide @ molten salt core-shell architecture as a generalized redox catalyst design strategy for chemical looping – oxidative dehydrogenation of ethane. 17 combinations of redox active oxides and molten salts were prepared, evaluated, and characterized. X-ray diffraction indicates that the redox oxides and molten salts are fully compatible, forming separate and stable phases. X-ray photoelectron spectroscopy demonstrates that the molten salts aggregate at the redox oxide surface, forming a core-shell structure to block the non-selective sites responsible for CO x formation. Up to ~74% single-pass olefin yields were achieved using the proposed redox catalyst design strategy. Statistical analyses of the performance data indicate the potential to achieve up to 86.7% single-pass yield by simply optimizing the operating conditions using the redox catalysts reported in this study. Meanwhile, the generalizability of the catalyst design strategy offers exciting opportunities to further optimize the composition and performance of the redox catalysts for ethane ODH under a chemical looping scheme with significantly reduced energy consumption and CO 2 emissions.

03 NATURAL GAS↗
DOE OSTI2021

A systematic comparison of machine learning methods for modeling of dynamic processes applied to combustion emission rate modeling

Ten established, data-driven dynamic algorithms are surveyed and a practical guide for understanding these methods generated. Existing Python programming packages for implementing each algorithm are acknowledged, and the model equations necessary for prediction are presented. A case study on a coal-fired power plant’s NO x emission rates is performed, directly comparing each modeling method’s performance on a mutual system. Each model is evaluated by its root mean squared error (RMSE) on out-of-sample future horizon predictions. Optimal hyperparameters are identified using either an exhaustive search or genetic algorithm. The top five model structures of each method are used to recursively predict future NO x emission rates over a 60-step time horizon. The RMSE at each future timestep is determined, and the recursive output prediction trends compared against measurements in time. The GRU neural network is identified as the best candidate for representing the system, demonstrating accurate and stable predictions across the future horizon by all considered models, while satisfactory performance was observed in several of the ARX/NARX formulations. Finally, these efforts have contributed 1) a concise resource of multiple proven dynamic machine learning methods, 2) a practical guide explaining the use of these methods, effectively lowering the “barrier-to-entry” of deploying such models in control systems, 3) a comparison study evaluating each method’s performance on a mutual system, 4) demonstration of accurate multi-timestep emissions modeling suitable for systems-level control, and 5) generalizable results demonstrating the suitability of each method for prediction over a multi-step future horizon to other complex dynamic systems.

42 ENGINEERING↗
DOE OSTI2026

Virtual refrigerant charge sensing algorithm for residential CO₂ heat pumps

Natural refrigerants are increasingly adopted in next-generation heat pump systems, among which CO₂ heat pumps have attracted significant attention. However, due to their high operating pressures, the leakage risk is higher, resulting in undercharge conditions and degraded heat pump performance. Thus, developing an accurate refrigerant charge level detection technique is necessary to guarantee safe and efficient operation. Although virtual refrigerant charge (VRC) level calculation algorithms for CO₂ heat pumps exist, they typically rely on empirically selected features without a systematic selection framework, leading to multicollinearity and potential overfitting, which limit their prediction accuracy and generalizability. To address these issues, this study proposes a VRC algorithm framework with a systematic feature selection method that identifies physically meaningful and statistically significant features, and is applied using a residential CO₂ heat pump as a case study. The method is extended from previous work on conventional refrigerants to account for charge behavior in CO₂ gas coolers. The selected features include gas cooler outlet density, evaporator pressure, and superheat temperature. The results demonstrate that the proposed feature selection method significantly improves prediction accuracy compared to existing VRC approaches. A relatively small training dataset (∼30 samples) is sufficient for feature identification and model development. The developed algorithm achieves less than 3% prediction error under both undercharge and overcharge conditions, representing reductions of 46.7% and 35.3% compared to two recent reference VRC algorithms for transcritical CO₂ heat pumps reported in the literature. The proposed algorithm and feature selection method enhance leakage detection capability, facilitate the deployment of CO₂ heat pump systems, and contribute to reduced energy waste and maintenance costs.

Guo, Fangzhou [Lawrence Berkeley National Laborato↗
DOE OSTI2020

Machine learning and ligand binding predictions: A review of data, methods, and obstacles

We report that computational predictions of ligand binding is a difficult problem, with more accurate methods being extremely computationally expensive. The use of machine learning for drug binding predictions could possibly leverage the use of biomedical big data in exchange for time-intensive simulations. This paper reviews current trends in the use of machine learning for drug binding predictions, data sources to develop machine learning algorithms, and potential problems that may lead to overfitting and ungeneralizable models. A few popular datasets that can be used to develop virtual high-throughput screening models are characterized using spatial statistics to quantify potential biases. We can see from evaluating some common benchmarks that good performance correlates with models with high-predicted bias scores and models with low bias scores do not have much predictive power. A better understanding of the limits of available data sources and how to fix them will lead to more generalizable models that will lead to novel drug discovery.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Electrochemical Characterization of Photo-Driven Hole-Scavenging by Cadmium Sulfide Quantum Dot-Nitrogenase Biohybrid Complexes

Light-driven biohybrid systems that couple semiconductor nanocrystals with enzymes offer a promising strategy for solar-to-chemical energy-conserving reduction reactions, yet are often limited by inefficient hole scavenging. Hole scavenging is critical for maintaining charge separation in the light-absorbing electron donor molecule and, thus, for sustaining catalytic turnover, making quantification essential to improving system efficiency. Here, we introduce a photoelectrochemical approach for real-time monitoring of hole scavenging in cadmium sulfide (CdS) quantum dot (QD)-nitrogenase MoFe protein biohybrids that catalyze the reduction of H+ and N2. Using hydroquinone (HQ) as the sacrificial electron donor (SED), oxidation of HQ to benzoquinone (BQ) by photogenerated valence-band holes is coupled to the electrochemical reduction of BQ at an electrode, enabling quantification of the hole-scavenging reaction via chronoamperometry. This approach provides a continuous, real-time readout of charge-transfer dynamics under illumination. Systematic variation of light intensity and donor concentration reveals that hole scavenging scales with photon flux and exhibits a non-linear dependence on SED concentration, with an optimum at intermediate HQ concentrations. These results establish chronoamperometry as a powerful tool for probing photoinduced charge transfer, advancing the quantitative understanding of hole-scavenging in biohybrid systems, and offering a generalizable framework for optimizing solar-driven biocatalysis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Data-driven equation-free dynamics applied to many-protein complexes: The microtubule tip relaxation

Microtubules (MTs) constitute the largest components of the eukaryotic cytoskeleton and play crucial roles in various cellular processes, including mitosis and intracellular transport. The property allowing MTs to cater to such diverse roles is attributed to dynamic instability, which is coupled to the hydrolysis of GTP (guanosine-5'-triphosphate) to GDP (guanosine-5'-diphosphate) within the β-tubulin monomers. Understanding the equilibrium dynamics and the structural features of both GDP- and GTP-complexed MT tips, especially at an all-atom level, remains challenging for both experimental and computational methods because of their dynamic nature and the prohibitive computational demands of simulating large, many-protein systems. This study employs the “equation-free” multiscale computational method to accelerate the relaxation of all-atom simulations of MT tips toward their putative equilibrium conformation. Using large MT lattice systems (14 protofilaments × 8 heterodimers) comprising ~21-38 million atoms, we applied this multiscale approach to leapfrog through time and nearly double the computational efficiency in realizing relaxed all-atom conformations of GDP- and GTP-complexed MT tips. Commencing from an initial 4 μs unbiased all-atom simulation, we interleave coarse projective “equation-free” jumps with short bursts of all-atom molecular dynamics simulation to realize an additional effective simulation time of 1.875 μs. Our 5.875 μs of effective simulation trajectories for each system expose the subtle yet essential differences in the structures of MT tips as a function of whether β-tubulin monomer is complexed with GDP or GTP, as well as the lateral interactions within the MT tip, offering a refined understanding of features underlying MT dynamic instability. Furthermore, the approach presents a robust and generalizable framework for future explorations of large biomolecular systems at atomic resolution.

Wu, Jiangbo [University of Chicago, IL (United Sta↗
DOE OSTI2023

Mesh refinement for anisotropic diffusion in magnetized plasmas

Highly accurate simulation of plasma transport is needed to drive the successful design and operation of magnetically confined fusion reactors. Unfortunately, the extreme anisotropy present in magnetized plasmas results in thin boundary layers that are expensive to resolve. Here, this work investigates how various mesh refinement strategies might reduce that expense to allow for more efficient simulation by comparing standard variable refinement approaches that use a field quantity to an adaptive approach that uses an error estimator. It is first verified that higher order discretization only realizes the proper rate of convergence once the mesh resolves the thin boundary layer, therefore motivating the focusing of refinement on the boundary layer. For three two-dimensional test cases that contain characteristic features of tokamak magnetic fields, an exponential refinement strategy based on the magnetic flux function, which is the standard refinement approach in the field, is compared to an adaptive strategy utilizing the established Zienwiekicz and Zhu error estimator. The adaptive mesh refinement strategy consistently achieves the same accuracy using orders of magnitude less degrees of freedom than either exponential or uniform refinement. This result makes the adaptive refinement strategy more efficient than the exponential refinement strategy while also being more generalizable to problems with complex magnetic geometries. Scaling laws are derived that quantify the improvement in cost of the adaptive refinement strategy over other refinement approaches in 2D and 3D.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Raspberry colloid-templated approach for the synthesis of palladium-based oxidation catalysts with enhanced hydrothermal stability and low-temperature activity

It is becoming increasingly urgent to develop and utilize novel, more efficient and stable materials for mobile and stationary emission control applications as the deleterious consequences of anthropogenic air pollution are becoming more evident and pressing. Tightening regulations, particularly related to automotive exhaust treatment, together with continued improvements in engine design, that result in lowering the engine operating temperatures and inadvertently lead to the release of an overwhelming proportion of pollutants during the cold start, present new challenges for materials design, specifically for oxidation catalysts. In particular, improvements in the low-temperature activity while maintaining catalyst stability at high temperatures are required from the next-generation catalyst. Typical catalysts for removal of pollutants from automotive exhaust streams incorporate platinum group metals (PGMs). They tend to be inefficient at low temperatures (below 250 °C), thus accounting for the cold start problem, yet sinter and lose their activity at high temperatures that are frequently encountered during catalyst operation. High PGM loadings are often employed to compensate for catalyst inefficiencies and fast degradation, ultimately resulting in high-cost catalytic converters. Here, we have developed a new approach for the design and formation of catalytic materials that allows for both significantly more efficient PGM incorporation and improved overall catalyst performance at reduced PGM loadings. The method provides control over the composition and geometry of the support through self-assembly of sacrificial composite template — “raspberry” polymeric colloids decorated with catalytic particles — accompanied by infiltration with metal-oxide precursor and subsequent removal of the colloids. This method simultaneously structures the porous network and organizes the catalytic particles within it. Uniquely, the resulting catalytic particles are partially embedded in the support matrix and partially exposed to the pore interior, producing catalytic sites that are both stable and accessible. Herein, the feasibility of this novel and versatile approach for automotive catalytic conversion is demonstrated: the studies include testing alumina-based raspberry-colloid-templated (RCT) catalysts containing Pd nanoparticles (RCT Pd/Al2O3) for oxidation of propane and carbon monoxide under simulated diesel exhaust conditions and hydrothermal aging at 800 °C for 50 h in the simulated stream. The RCT Pd/Al2O3 catalysts exhibit exceptional activity toward CO oxidation, reduced reaction onset temperature, and high stability to elevated temperatures (demonstrated through prolonged exposure to temperatures up to 950 °C) and reactive gas streams, without migration, sintering or loss of the precious metal NPs. Notably, the novel catalyst shows the same or slightly better performance than the commercial catalysts even when the PGM load is reduced by ~80 % compared to the commercial counterparts. These results provide confidence for the utilization of the RCT approach for the fabrication of robust nanostructured catalysts for next-generation, energy-efficient catalytic converters with improved performance at low and high temperatures and reduced costs. The RCT methodology is, in addition, highly generalizable, and can thus be applied for the design of a wide range of catalytic systems in the automotive sector and beyond.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Mitigation of chromosome loss in clinical CRISPR-Cas9-engineered T cells

CRISPR-Cas9 genome editing has enabled advanced T cell therapies, but occasional loss of the targeted chromosome remains a safety concern. To investigate whether Cas9-induced chromosome loss is a universal phenomenon and evaluate its clinical significance, we conducted a systematic analysis in primary human T cells. Arrayed and pooled CRISPR screens revealed that chromosome loss was generalizable across the genome and resulted in partial and entire loss of the targeted chromosome, including in preclinical chimeric antigen receptor T cells. T cells with chromosome loss persisted for weeks in culture, implying the potential to interfere with clinical use. A modified cell manufacturing process, employed in our first-in-human clinical trial of Cas9-engineered T cells (NCT03399448), reduced chromosome loss while largely preserving genome editing efficacy. Expression of p53 correlated with protection from chromosome loss observed in this protocol, suggesting both a mechanism and strategy for T cell engineering that mitigates this genotoxicity in the clinic.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

The unique architecture of umbrella toxins permits a two-tiered molecular bet-hedging strategy for interbacterial antagonism

Bacteria exist in competitive and rapidly changing environments in which the nature of future threats cannot be easily predicted. Streptomyces coelicolor produces three antibacterial umbrella particles that harbor distinct polymorphic toxin domains and an overlapping set of six diversified lectins. Here, we show that the exquisite specificity of umbrella particles derives from lectin-mediated species-specific binding to previously undescribed hypervariable surface glycoconjugates. A cryo-electron microscopy (cryo-EM) structure of one such lectin in complex with its oligosaccharide substrate defines the molecular basis for targeting through the coordinated recognition of multiple glycan features. Biochemical and genetic studies of several target species, in conjunction with lectin-swapping experiments, support a model whereby S. coelicolor umbrella toxin diversification at the levels of lectin composition and toxin polymorphism represents a unique, two-tiered bet-hedging strategy. Bioinformatic analyses support this as a means by which the unusual architecture of umbrella toxins offers Streptomyces a generalizable strategy to antagonize an unpredictable array of competitors.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

A high-throughput workflow to analyze sequence-conformation relationships and explore hydrophobic patterning in disordered peptoids

Understanding how a macromolecule’s primary sequence governs its conformational landscape is crucial for elucidating its function, yet these design principles are still emerging for macromolecules with intrinsic disorder. Herein, we introduce a high-throughput workflow that implements a practical colorimetric conformational assay, introduces a semi-automated sequencing protocol using matrix-assisted laser desorption/ionization and tandem mass spectrometry (MALDI-MS/MS), and develops a generalizable sequence-structure algorithm. Using a model system of 20mer peptidomimetics containing polar glycine and hydrophobic N-butylglycine residues, we identified nine classifications of conformational disorder and isolated 122 unique sequences across varied compositions and conformations. Conformational distributions of three compositionally identical library sequences were corroborated through atomistic simulations and ion mobility spectrometry coupled with liquid chromatography. A data-driven strategy was developed using existing sequence variables and data-derived “motifs” to inform a machine-learning algorithm toward conformation prediction. Here, this multifaceted approach enhances our understanding of sequence-conformation relationships and offers a powerful tool for accelerating the discovery of materials with conformational control.

data-driven analysis↗
DOE OSTI2024

Peridynamic neural operators: A data-driven nonlocal constitutive model for complex material responses

Neural operators, which can act as implicit solution operators of hidden governing equations, have recently become popular tools for learning the responses of complex real-world physical systems. Nevertheless, most neural operator applications have thus far been data-driven and neglect the intrinsic preservation of fundamental physical laws in data. Here, in this work, we introduce a novel integral neural operator architecture called the Peridynamic Neural Operator (PNO) that learns a nonlocal constitutive law from data. This neural operator provides a forward model in the form of state-based peridynamics, with objectivity and momentum balance laws automatically guaranteed. As applications, we demonstrate the expressivity and efficacy of our model in learning complex material behaviors from both synthetic and experimental data sets. We also compare the performances with baseline models that use predefined constitutive laws. We show that, owing to its ability to capture complex responses, our learned neural operator achieves improved accuracy and efficiency. Moreover, by preserving the essential physical laws within the neural network architecture, the PNO is robust in treating noisy data. The method shows generalizability to different domain configurations, external loadings, and discretizations.

42 ENGINEERING↗
DOE OSTI2025

Physics-informed latent neural operator for real-time predictions of time-dependent parametric PDEs

Deep operator network (DeepONet) has shown significant promise as surrogate models for systems governed by partial differential equations (PDEs), enabling accurate mappings between infinite-dimensional function spaces. However, when applied to systems with high-dimensional input-output mappings arising from large numbers of spatial and temporal collocation points, these models often require heavily overparameterized networks, leading to long training times. Latent DeepONet addresses some of these challenges by introducing a two-step approach: first learning a reduced latent space using a separate model, followed by operator learning within this latent space. While efficient, this method is inherently data-driven and lacks mechanisms for incorporating physical laws, limiting its robustness and generalizability in data-scarce settings. Here, in this work, we propose PI-Latent-NO, a physics-informed latent neural operator framework that integrates governing physics directly into the learning process. Our architecture features two coupled DeepONets trained end-to-end: a Latent-DeepONet that learns a low-dimensional representation of the solution, and a Reconstruction-DeepONet that maps this latent representation back to the physical space. By embedding PDE constraints into the training via automatic differentiation, our method eliminates the need for labeled training data and ensures physics-consistent predictions. The proposed framework is both memory and compute-efficient, exhibiting near-constant scaling with problem size and demonstrating significant speedups over traditional physics-informed operator models. We validate our approach on a range of parametric PDEs, showcasing its accuracy, scalability, and suitability for real-time prediction in complex physical systems.

Latent representations↗