SEARCH · Engineering Papers

Results for “collocated”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 199 records · Page 11

All sources NASA NTRS DOE OSTI

NASA NTRS2005

Computation of Sound Propagation by Boundary Element Method

This report documents the development of a Boundary Element Method (BEM) code for the computation of sound propagation in uniform mean flows. The basic formulation and implementation follow the standard BEM methodology; the convective wave equation and the boundary conditions on the surfaces of the bodies in the flow are formulated into an integral equation and the method of collocation is used to discretize this equation into a matrix equation to be solved numerically. New features discussed here include the formulation of the additional terms due to the effects of the mean flow and the treatment of the numerical singularities in the implementation by the method of collocation. The effects of mean flows introduce terms in the integral equation that contain the gradients of the unknown, which is undesirable if the gradients are treated as additional unknowns, greatly increasing the sizes of the matrix equation, or if numerical differentiation is used to approximate the gradients, introducing numerical error in the computation. It is shown that these terms can be reformulated in terms of the unknown itself, making the integral equation very similar to the case without mean flows and simple for numerical implementation. To avoid asymptotic analysis in the treatment of numerical singularities in the method of collocation, as is conventionally done, we perform the surface integrations in the integral equation by using sub-triangles so that the field point never coincide with the evaluation points on the surfaces. This simplifies the formulation and greatly facilitates the implementation. To validate the method and the code, three canonic problems are studied. They are respectively the sound scattering by a sphere, the sound reflection by a plate in uniform mean flows and the sound propagation over a hump of irregular shape in uniform flows. The first two have analytical solutions and the third is solved by the method of Computational Aeroacoustics (CAA), all of which are used to compare the BEM solutions. The comparisons show very good agreements and validate the accuracy of the BEM approach implemented here.

Guo, Yueping↗

NASA NTRS2011

Fast Simulators for Satellite Cloud Optical Centroid Pressure Retrievals, 1. Evaluation of OMI Cloud Retrievals

We have developed a relatively simple scheme for simulating retrieved cloud optical centroid pressures (OCP) from satellite solar backscatter observations. We have compared simulator results with those from more detailed retrieval simulators that more fully account for the complex radiative transfer in a cloudy atmosphere. We used this fast simulator to conduct a comprehensive evaluation of cloud OCPs from the two OMI algorithms using collocated data from CloudSat and Aqua MODIS, a unique situation afforded by the A-train formation of satellites. We find that both OMI algorithms perform reasonably well and that the two algorithms agree better with each other than either does with the collocated CloudSat data. This indicates that patchy snow/ice, cloud 3D, and aerosol effects not simulated with the CloudSat data are affecting both algorithms similarly. We note that the collocation with CloudSat occurs mainly on the East side of OMI's swath. Therefore, we are not able to address cross-track biases in OMI cloud OCP retrievals. Our fast simulator may also be used to simulate cloud OCP from output generated by general circulation models (GCM) with appropriate account of cloud overlap. We have implemented such a scheme and plan to compare OMI data with GCM output in the near future.

Joiner, J.↗

NASA NTRS2013

High resolution wind measurements for offshore wind energy development

A method, apparatus, system, article of manufacture, and computer readable storage medium provide the ability to measure wind. Data at a first resolution (i.e., low resolution data) is collected by a satellite scatterometer. Thin slices of the data are determined. A collocation of the data slices are determined at each grid cell center to obtain ensembles of collocated data slices. Each ensemble of collocated data slices is decomposed into a mean part and a fluctuating part. The data is reconstructed at a second resolution from the mean part and a residue of the fluctuating part. A wind measurement is determined from the data at the second resolution using a wind model function. A description of the wind measurement is output.

Nghiem, Son Van↗

NASA NTRS2013

Coherent Uncertainty Analysis of Aerosol Measurements from Multiple Satellite Sensors

Aerosol retrievals from multiple spaceborne sensors, including MODIS (on Terra and Aqua), MISR, OMI, POLDER, CALIOP, and SeaWiFS altogether, a total of 11 different aerosol products were comparatively analyzed using data collocated with ground-based aerosol observations from the Aerosol Robotic Network (AERONET) stations within the Multi-sensor Aerosol Products Sampling System (MAPSS, http://giovanni.gsfc.nasa.gov/mapss/ and http://giovanni.gsfc.nasa.gov/aerostat/). The analysis was performed by comparing quality-screened satellite aerosol optical depth or thickness (AOD or AOT) retrievals during 2006-2010 to available collocated AERONET measurements globally, regionally, and seasonally, and deriving a number of statistical measures of accuracy. We used a robust statistical approach to detect and remove possible outliers in the collocated data that can bias the results of the analysis. Overall, the proportion of outliers in each of the quality-screened AOD products was within 12%. Squared correlation coefficient (R2) values of the satellite AOD retrievals relative to AERONET exceeded 0.6, with R2 for most of the products exceeding 0.7 over land and 0.8 over ocean. Root mean square error (RMSE) values for most of the AOD products were within 0.15 over land and 0.09 over ocean. We have been able to generate global maps showing regions where the different products present advantages over the others, as well as the relative performance of each product over different landcover types. It was observed that while MODIS, MISR, and SeaWiFS provide accurate retrievals over most of the landcover types, multi-angle capabilities make MISR the only sensor to retrieve reliable AOD over barren and snow / ice surfaces. Likewise, active sensing enables CALIOP to retrieve aerosol properties over bright-surface shrublands more accurately than the other sensors, while POLDER, which is the only one of the sensors capable of measuring polarized aerosols, outperforms other sensors in certain smoke-dominated regions, including broadleaf evergreens in Brazil and South-East Asia.

remote sensing↗

NASA NTRS2013

Integrated High-Speed Torque Control System for a Robotic Joint

A control system for achieving high-speed torque for a joint of a robot includes a printed circuit board assembly (PCBA) having a collocated joint processor and high-speed communication bus. The PCBA may also include a power inverter module (PIM) and local sensor conditioning electronics (SCE) for processing sensor data from one or more motor position sensors. Torque control of a motor of the joint is provided via the PCBA as a high-speed torque loop. Each joint processor may be embedded within or collocated with the robotic joint being controlled. Collocation of the joint processor, PIM, and high-speed bus may increase noise immunity of the control system, and the localized processing of sensor data from the joint motor at the joint level may minimize bus cabling to and from each control node. The joint processor may include a field programmable gate array (FPGA).

Davis, Donald R.↗

NASA NTRS2016

Parallel Aircraft Trajectory Optimization with Analytic Derivatives

Trajectory optimization is an integral component for the design of aerospace vehicles, but emerging aircraft technologies have introduced new demands on trajectory analysis that current tools are not well suited to address. Designing aircraft with technologies such as hybrid electric propulsion and morphing wings requires consideration of the operational behavior as well as the physical design characteristics of the aircraft. The addition of operational variables can dramatically increase the number of design variables which motivates the use of gradient based optimization with analytic derivatives to solve the larger optimization problems. In this work we develop an aircraft trajectory analysis tool using a Legendre-Gauss-Lobatto based collocation scheme, providing analytic derivatives via the OpenMDAO multidisciplinary optimization framework. This collocation method uses an implicit time integration scheme that provides a high degree of sparsity and thus several potential options for parallelization. The performance of the new implementation was investigated via a series of single and multi-trajectory optimizations using a combination of parallel computing and constraint aggregation. The computational performance results show that in order to take full advantage of the sparsity in the problem it is vital to parallelize both the non-linear analysis evaluations and the derivative computations themselves. The constraint aggregation results showed a significant numerical challenge due to difficulty in achieving tight convergence tolerances. Overall, the results demonstrate the value of applying analytic derivatives to trajectory optimization problems and lay the foundation for future application of this collocation based method to the design of aircraft with where operational scheduling of technologies is key to achieving good performance.

aircraft↗

NASA NTRS2017

Implicit Formulation of Muscle Dynamics in OpenSim

Astronauts lose bone and muscle mass during spaceflight. Exercise countermeasure is the primary method for counteracting bone and muscle mass loss in space. New spacecraft exercise device concepts are currently being developed for the NASAs new crew exploration vehicle. The NASA Digital Astronaut Project (DAP) uses computational modeling to help determine if the new exercise devices will be effective as countermeasures. The NASA Digital Astronaut Project is developing the ability to utilize predictive simulation to provide insight into the change in kinematics and kinetics with a change in device and gravitational environment (1-g versus 0-g). For example, in space exercise the subject's body weight is applied in addition to the loads prescribed for musculoskeletal maintenance. How and where these loads are applied obviously directly impacts bone and tissue loads. Additionally, due to space vehicle structural requirements, exercise devices are often placed on vibration isolation systems. This changes the apparent impedance or stiffness of the device as seen by the user. Data collection under these conditions is often impractical and limited. Predictive modeling provides a means to have a virtual subject to test hypotheses. Predictive simulation provides a virtual subject for which we are able to perform studies such as sensitivity to device loading and vibration isolation without the need for laboratory kinematic or kinetic test data.Direct Collocation optimization provides an efficient means to perform task based optimization and predictive modeling. It is relatively straight forward to structure a physical exercise task in a Direct Collocation mathematical formulation: perform a motion such that you start at an initial pose, achieve a given amount of deflection i.e a squat, return to the initial pose, and minimize muscle activation cost. Direct Collocation is advantageous in that it does not require numerical integration to evaluate the objective function. Instead, the system dynamics are transformed to discrete time and the optimizer is constrained such that the solution is not considered to be a valid unless the dynamic equations are satisfied at all time points. The simulation and optimization are effectively done simultaneously. Due to the implicit integration, time steps can be more coarse than in a differential equation solver. In a gait scenario this means that that the model constraints and cost function are evaluated at 100 nodes in the gait cycle versus 10,000 integration steps in a variable-step forward dynamic simulation. Furthermore, no time is wasted on accurate simulations of movements that are far from the optimum. Constrained optimization algorithms require a Jacobian matrix that contains the partial derivatives of each of the dynamic constraints with respect to of each of the state and control variables at all time points. This is a large but sparse matrix. An implicit dynamics formulation requires computation of the dynamic residuals f as a function of the states x and their derivatives, and controls u:f(x, dxdt, u) 0If the dynamics of musculoskeletal system are formulated implicitly, the Jacobian elements are often available analytically, eliminating the need for numerical differentiation; this is obviously computationally advantageous. Additionally, implicit formulation of musculoskeletal dynamics do not suffer from singularities from low mass bodies, zero muscle activation, or other stiff system or

physical exercise↗

NASA NTRS2019

Low-Thrust Trajectory Design for a Cislunar CubeSat Leveraging Structures from the Bicircular Restricted Four-Body Problem

The upcoming Lunar IceCube (LIC) mission will deliver a 6U CubeSat to a low lunar orbit via a ride-share opportunity during NASAs Artemis 1 mission. This presents a challenging trajectory design scenario, as the vast change in energy required to transfer from the initial deployment state to the destination orbit is compounded by the limitations of the LICs low-thrust engine. This investigation addresses these challenges by developing a trajectory design framework that utilizes dynamical structures available in the Bicircular Restricted Four-Body Problem (BCR4BP) along with a robust direct collocation algorithm. Maps are created that expedite the selection of invariant manifold paths from a periodic staging orbit in the BCR4BP that offer favorable connections between the LIC transfer phases. Initial guesses assembled from these maps are passed to a direct collocation algorithm that corrects them in the BCR4BP while including the variable low-thrust acceleration of the spacecraft engine. Results indicate that the ordered motion provided by the BCR4BP and the robustness of direct collocation combine to offer an efficient and adaptable framework for designing a baseline trajectory for the LIC mission.

Pritchett, Robert↗

NASA NTRS2019

Low-Thrust Trajectory Design for A Cislunar CubeSat Leveraging Structures from the Bicircular Restricted Four-Body Problem

Pritchett, Robert↗

NASA NTRS2020

A Machine Learning-Based Cloud Detection and Thermodynamic Phase Classification Algorithm using Passive Spectral Observations

We trained two Random Forest (RF) machine-learning models for cloud mask and cloud thermodynamic phase detection using spectral observations from VIIRS on Suomi NPP (SNPP). Observations from CALIOP were carefully selected to provide reference labels. The two RF models were trained for all-day and daytime-only conditions using a 4-year collocated VIIRS/CALIOP dataset from 2013 to 2016. Due to the orbit difference, the collocated CALIOP and SNPP VIIRS training samples cover a broad viewing zenith angle range, which is a great benefit to overall model performance. The all-day model uses 3 VIIRS infrared (IR) bands (8.6,11, and 12 μm) and the daytime model uses 5 Near-IR (NIR) and Shortwave-IR (SWIR) bands (0.86, 1.24, 1.38, 1.64 and 2.25 μm) together with the 3 IR bands to detect clear, liquid water, and ice cloud pixels. Up to 7 surface types, namely, ocean/water, forest, cropland, grassland, snow/ice, barren/desert, and shrubland, were considered separately to enhance performance for both models. Detection of cloudy pixels and thermodynamic phase with the two RF models were compared against collocated CALIOP products from 2017. It is shown that, with a conservative screening process that excludes the most challenging cloudy pixels for passive remote sensing, the two RF models have high accuracy rates in comparison with the CALIOP reference for both cloud detection and thermodynamic phase. Other existing SNPP VIIRS and Aqua MODIS cloud mask and phase products are also evaluated, with results showing that the two RF models and the MODIS MYD06 optical property phase product are the top 3 algorithms with respect to lidar observations during the daytime. During the nighttime, the RF all-day model works best for both cloud detection and phase, in particular for pixels over snow/ice surfaces. The present RF models can be extended to other similar passive instruments if training samples can be collected from CALIOP or other lidars. However, the quality of reference labels and potential sampling issues that may impact model performance would need further attention.

cloud detection↗

NASA NTRSDate not supplied

Discontinuous Galerkin and Related Methods for ODE

A defining feature of the discontinuous Galerkin (DG) method for ODE is that the piecewise polynomial solution can have a jump discontinuity at the beginning of each step. Starting from the standard integral formulation, the DG method is derived here in differential form. The key ingredient is a polynomial called the correction function, which helps ‘correct’ the discontinuous solution by approximating the jump and yields a continuous one. Under the right Radau quadrature, this continuous solution is identical to the solutions by the right Radau collocation and the continuous Galerkin (CG) methods. Next, the correction function facilitates the construction of the associated implicit Runge-Kutta schemes (IRK-DG). Different quadratures for DG result in different IRK-DG methods: left Radau quadrature in Radau IA, right Radau quadrature in Radau IIA or right Radau collocation, and Gauss quadrature in a method called DG-Gauss. The construction of IRK-DG clarifies the meaning and facilitates the proofs of various 𝐵(𝑝), 𝐶(𝜂), and 𝐷(𝜁) conditions for accuracy. The two consequences of these conditions are that all 𝑠-stage IRK-DG methods are accurate to order 2𝑠 − 1, and the IRK-DG methods of Radau type are unique. Numerical examples showing the behavior of the DG solutions are provided. In all, the correction function plays a key role and helps establish the relations among the DG, IRK-DG, collocation, and CG schemes.

numerical methods↗

NASA NTRSDate not supplied

Discontinuous Galerkin and Related Methods for ODE

Starting from the standard integral formulation, the DG method is derived here in differential form. The key ingredient is a polynomial called the correction function, which helps ‘correct’ the discontinuous solution by approximating the jump and yields a continuous one. Under the right Radau quadrature, this continuous solution is identical to the solutions by the right Radau collocation and the continuous Galerkin (CG) methods. Next, the correction function facilitates the construction of the associated implicit Runge-Kutta schemes (IRK-DG). Different quadratures for DG result in different IRK-DG methods: left Radau quadrature in Radau IA, right Radau quadrature in Radau IIA or right Radau collocation, and Gauss quadrature in a method called DG-Gauss. The construction of IRK-DG clarifies the meaning and facilitates the proofs of various 𝐵(𝑝), 𝐶(𝜂), and 𝐷(𝜁) conditions for accuracy. The two consequences of these conditions are that all 𝑠-stage IRK-DG methods are accurate to order 2𝑠− 1, and the IRK-DG methods of Radau type are unique. Numerical examples showing the behavior of the DG solutions are provided. In all, the correction function plays a key role and helps establish the relations among the DG, IRK DG, collocation, and CG methods.

Numerical Methods for Ordinary Differential Equati↗

NASA NTRSDate not supplied

Creating Seasonal Climatologies of Aerosol Lidar Ratios Over Ocean Using MODIS AOD Constrained Retrievals and GOCART Model Simulations

The current CALIPSO algorithms assign one lidar ratio (i.e., extinction-to-backscatter ratio; LR) value globally for each of the seven tropospheric aerosol types. In a future data products release, the CALIPSO project aims to improve these algorithms through the development of regional and seasonal LR climatologies. In this work, aerosol LRs are inferred through CALIOP backscatter profiles constrained by collocated aerosol optical depth (AOD) from Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) data. This analysis is subsampled for those profiles that are cloud-free and contain only one CALIOP-classified aerosol type (e.g., marine). The CALIOP profiles are then collocated with aerosol volume fractions obtained through Goddard Chemistry Aerosol Radiation and Transport (GOCART) model simulations. In this poster, we show twelve-year (June 2006-August 2018) mean spatial distributions of inferred aerosol LRs for CALIOP-classified marine aerosols and how they correspond inversely with patterns of GOCART sea salt volume fraction (SSVF). Near land masses, smaller SSVFs and larger LRs are found (due to the influence of over-land aerosols). In the remote ocean regions (likely less impacted by over-land aerosols), larger SSVFs and smaller LRs are found. The developed relationship between the GOCART model SSVFs and MODIS AOD constrained LRs is used to create model-assisted seasonal LR maps. Additionally, we show maps of inferred LRs from constrained retrievals using the CALIPSO Ocean Derived Column Optical Depth (ODCOD) product and comparisons with those from the MODIS analyses. The technique demonstrated here benefits CALIPSO in the near-term, but similar methods can also be applied to the next generation space-based elastic backscatter lidars with collocated passive sensors, such as those of the upcoming NASA Atmosphere Observing System (AOS).

Travis Toth↗

NASA NTRSDate not supplied

Mapping Aerosol Lidar Ratios Over Ocean Using Constrained Retrievals and A Global Aerosol Model

The current NASA Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) algorithms assign one lidar ratio (i.e., extinction-to-backscatter ratio; LR) value globally for each of the seven tropospheric aerosol types. In a future data products release, the CALIPSO project aims to improve these algorithms through the development of regional and seasonal LR climatologies. In this work, aerosol LRs are inferred through Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) backscatter profiles constrained by collocated aerosol optical depth (AOD) from Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) data. This analysis is subsampled for those profiles that are cloud-free and contain only one CALIOP-classified aerosol type (e.g., marine). The CALIOP profiles are then collocated with aerosol volume fractions obtained through Goddard Chemistry Aerosol Radiation and Transport (GOCART) model simulations. In this presentation, we show that the twelve-year (June 2006-August 2018) mean spatial distributions of inferred aerosol LRs for CALIOP-classified marine aerosols correspond inversely with patterns of GOCART sea salt volume fraction (SSVF). For example, smaller SSVFs (< 65%) and larger LRs (> 55 sr), are found near land masses. This is indicative of the influence of over-land aerosols (e.g., pollution and biomass burning smoke). In the remote oceans (i.e., regions likely less impacted by non-sea salt aerosols), larger SSVFs (> 95%) and smaller LRs (< 25 sr) are found. The developed relationship between the GOCART SSVFs and MODIS AOD constrained LRs (polynomial fit intersect values of ~56 sr for SSVF of 0% and ~21 sr for SSVF of 100%) is used to produce model-assisted climatological LR maps on seasonal scales. Additionally, we show maps of inferred LRs from constrained retrievals using the CALIPSO Ocean Derived Column Optical Depth (ODCOD) product and comparisons with those from the MODIS analyses. The technique demonstrated in this study not only benefits CALIPSO in the near-term, but similar methods can be applied to future spaceborne elastic backscatter lidars with collocated passive sensors, such as those associated with the upcoming NASA Atmosphere Observing System (AOS).

Travis Toth↗

NASA NTRSDate not supplied

Mapping Aerosol Lidar Ratios Over Ocean using MODIS AOD Constrained Retrievals and GOCART Model Simulations

After 17 years, the NASA Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) mission ceased science operations in August 2023. For the final CALIPSO data products release (Version 5), the CALIPSO project seeks to improve the accuracy of its aerosol extinction by advancing knowledge of aerosol lidar ratios (i.e., extinction-to-backscatter ratios; LRs) for various aerosol types. The current algorithm assigns one LR value globally for each of the seven tropospheric aerosol types. The CALIPSO team aims to improve the retrieval algorithm through the development of regional and seasonal LR climatologies for the same aerosol types. In this study, aerosol LRs are inferred through Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) backscatter profiles constrained by collocated aerosol optical depth (AOD) from Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) data over oceans during daytime. This analysis is subsampled for those profiles that are cloud-free and contain only one CALIOP-classified aerosol type. The CALIOP profiles are then collocated with aerosol volume fractions obtained through Goddard Chemistry Aerosol Radiation and Transport (GOCART) model simulations. This presentation will reveal findings that the 12-year (June 2006-August 2018) mean spatial distributions of inferred aerosol LRs for CALIOP-classified marine and dusty marine aerosols correspond inversely with patterns of GOCART sea salt volume fraction (SSVF). For example, smaller SSVFs (< 65%) and larger LRs (> 55 sr), are found near land masses (Fig. 1). This indicates the influence of advected anthropogenic aerosols (e.g., pollution and biomass burning smoke). In the remote oceans (i.e., regions likely less impacted by non-sea salt aerosols), the SSVFs are larger (> 95%) and the LRs are smaller (< 25 sr) (Fig. 1). A polynomial fit of the MODIS AOD constrained LRs to the corresponding GOCART SSVFs (intersect values of ~58 sr for SSVF of 0% and ~21 sr for SSVF of 100%) is further used to produce model-assisted climatological LR maps on seasonal scales. Additionally, we will show results of a LR validation analysis for which we compare the revised CALIPSO AODs obtained by applying the seasonal/regional constrained LRs against CALIPSO Version 4.51 Ocean Derived Column Optical Depth (ODCOD). While the majority of the presentation will focus on LRs for CALIOP-classified marine and dusty marine aerosols, an overview of LR results will show preliminary results for other aerosol types over ocean, such as dust and elevated smoke. The technique demon-strated in this study highlights the benefits not only to the final planned CALIPSO data release in 2025, but similar methods can be applied to future spaceborne elastic backscatter lidars with collocated passive sensors (e.g., such as those associated with NASA’s proposed Atmosphere Observing System).

Travis D Toth↗

DOE OSTI2022

Evaluating Grid Strength under Uncertain Renewable Generation

The increasing displacement of synchronous generators with renewable resources such as wind and solar via power electronic interfaces causes a reduction in short-circuit strength and weak grid issues. The variation and uncertainty of renewable energy increase challenges for identifying weak grid conditions. This paper proposes an efficient method to analyze the impact of uncertain renewable energy on grid strength. The proposed method uses the probabilistic collocation method (PCM) to approximate the results of grid strength assessment under uncertain renewable generation, in order to reduce computational burden without compromising result accuracy when compared with traditional Monte Carlo simulation (MCS). To improve the accuracy of the approximation results, the proposed method integrates the K-means clustering technique with PCM to select the approximation samples of input variables. The efficacy of the proposed method is demonstrated by comparison with MCS on the modified IEEE 9-bus system and modified IEEE 39-bus system with multiple renewable generators.

grid strength↗

DOE OSTI2025

Constrained or unconstrained? Neural-network-based equation discovery from data

Throughout many fields, practitioners often rely on differential equations to model systems. Yet, for many applications, the theoretical derivation of such equations and/or the accurate resolution of their solutions may be intractable. Instead, recently developed methods, including those based on parameter estimation, operator subset selection, and neural networks, allow for the data-driven discovery of both ordinary and partial differential equations (PDEs), on a spectrum of interpretability. The success of these strategies is often contingent upon the correct identification of representative equations from noisy observations of state variables and, as importantly and intertwined with that, the mathematical strategies utilized to enforce those equations. Specifically, the latter has been commonly addressed via unconstrained optimization strategies. Representing the PDE as a neural network, we propose to discover the PDE (or the associated operator) by solving a constrained optimization problem and using an intermediate state representation similar to a physics-informed neural network (PINN). The objective function of this constrained optimization problem promotes matching the data, while the constraints require that the discovered PDE is satisfied at a number of spatial collocation points. We present a penalty method and a widely used trust-region barrier method to solve this constrained optimization problem, and we compare these methods on numerical examples. Our results on several example problems demonstrate that the latter constrained method outperforms the penalty method, particularly for higher noise levels or fewer collocation points. This work motivates further exploration into using sophisticated constrained optimization methods in scientific machine learning, as opposed to their commonly used, penalty-method or unconstrained counterparts. For both of these methods, we solve these discovered neural network PDEs with classical methods, such as finite difference methods, as opposed to PINNs-type methods relying on automatic differentiation. Here, we briefly highlight how simultaneously fitting the data while discovering the PDE improves the robustness to noise and other small, yet crucial, implementation details.

Data-driven discovery↗

DOE OSTI2025

Gradient flow based phase-field modeling using separable neural networks

Allen–Cahn equation is a reaction–diffusion equation and is widely used for modeling phase separation. Machine learning methods for solving the Allen–Cahn equation in its strong form suffer from inaccuracies in collocation techniques, errors in computing higher-order spatial derivatives, and the large system size required by the space–time approach. To overcome these challenges, we propose solving the gradient flow of the Ginzburg–Landau free energy functional, which is equivalent to the Allen–Cahn equation, thereby avoiding the second-order spatial derivatives associated with the Allen–Cahn equation. A minimizing movement scheme is employed to solve the gradient flow problem, eliminating the complexities of a space–time approach. We utilize a separable neural network that efficiently represents the phase field through low-rank tensor decomposition. As we use the minimizing movement scheme to numerically solve the gradient flow problem, we thus, refer to the proposed method as the Separable Deep Minimizing Movement (SDMM) method. The evaluation of the functional in the minimizing movement scheme using the Gauss quadrature technique bypasses the inaccuracies associated with collocation techniques traditionally used to solve partial differential equations. A hyperbolic tangent transformation is introduced on the phase field prior to the evaluation of the functional to ensure that it remains strictly bounded within the values of the two phases. For this transformation, theoretical guarantee for energy stability of the minimizing movement scheme is established. Our results suggest that this transformation helps to improve the accuracy and efficiency significantly. The proposed method resolves the challenges faced by state-of-the-art machine learning techniques, outperforming them in both accuracy and efficiency. It is also the first machine learning method to achieve an order of magnitude speed improvement over the finite element method. In addition to its formulation and computational implementation, several case studies illustrate the applicability of the proposed method.

42 ENGINEERING↗