Dimensionality Reduction and Weight-Parameterized Neural NetworkSurrogates for Climate Models
Explore the source record for details and available documents.
Diaz-Ibarra, Oscar Homero↗
SEARCH · Engineering Papers
Results for “model weights”
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.
Physics-guided Deep Learning for Power System State Estimation
In the past decade, dramatic progress has been made in the field of machine learning. This paper explores the possibility of applying deep learning in power system state estimation. Traditionally, physics-based models are used including weighted least square (WLS) or weighted least absolute value (WLAV). These models typically consider a single snapshot of the system without capturing temporal correlations of system states. In this paper, a physics-guided deep learning (PGDL) method is proposed. Specifically, inspired by autoencoders, deep neural networks (DNNs) are used to learn the temporal correlations. The estimated system states from DNNs are then checked against physics laws by running through a set of power flow equations. Hence, the proposed PGDL is both data-driven and physics-guided. The accuracy and robustness of the proposed PGDL method are compared with traditional methods in standard IEEE cases. Simulations show promising results and the applicability is further discussed.
24 POWER TRANSMISSION AND DISTRIBUTION↗
Intumescent coating modeling
A thermal response mathematical model for an intumescent coating system is developed. The model predicts indepth temperature profiles of the solid as a function of the fundamental kinetic, mass, and heat-transfer properties of the original coating system. In the model, swelling is based on an empirical function of the paint's weight loss. The model is used to describe the effectiveness of the coating for thermal protection. Using the thermal response time at the backface of the material as a criterion, a sensitivity analysis is made with respect to variation in individual properties, such as total expansion, to provide a guide for designing improved intumescing agents. The sensitivity analysis shows that total expansion and rate of expansion, along with other material properties, are important parameters for providing thermal protection. Finally, backface temperature histories are predicted and compared to experimental data obtained from a representative fire environment. Predicted temperatures agree within a maximum error of plus or minus 20%.
Cagliostro, D. E.↗
Determination of ocean tides from the first year of TOPEX/POSEIDON altimeter measurements
An improved geocentric global ocean tide model has been determined using 1 year of TOPEX/POSEIDON altimeter measurements to provide corrections to the Cartwright and Ray (1991) model (CR91). The corrections were determined on a 3 deg x 3 deg grid using both the harmonic analysis method and the response method. The two approaches produce similar solutions. The effect on the tide solution of simultaneously adjusting radial orbit correction parameters using altimeter measurements was examined. Four semidiurnal (N(sub 2), M(sub 2), S(sub 2) and K(sub 2)), four diurnal (Q(sdub 1), O(sub 1), P(sub 1), and K(sub 1)), and three long-period (S(sub sa), M(sub m), and M(sub f)) constituents, along with the variations at the annual frequency, were included in the harmomnic analysis solution. The observed annual variations represents the first global measurement describing accurate seasonal changes of the ocean during an El Nino year. The corrections to the M(sub 2) constituent have an root mean square (RMS) of 3.6 cm and display a clear banding pattern with regional highs and lows reaching 8 cm. The improved tide model reduces the weighted altimeter crossover residual from 9.8 cm RMS, when the CR91 tide model is used, to 8.2 cm on RMS. Comparison of the improved model to pelagic tidal constants determined from 80 tide gauges gives RMS differences of 2.7 cm for M(sub 2) and 1.7 cm for K(sub 1). Comparable values when the CR91 model is used are 3.9 cm and 2.0 cm, respectively. Examination of TOPEX/POSEIDON sea level anomaly variations using the new tide model further confirms that the tide model has been improved.
Ma, X. C.↗
An improved lattice Boltzmann scheme for multiphase fluid with multi-range interactions
Modeling of fluids with liquid to gas phase transition has become important for understanding many environmental or industrial processes. Such simulations need new techniques, because traditional solvers are often limited. The Lattice Boltzmann Model (LBM) allows simulate complex fluids, because its mesoscopic nature gives possibility to incorporate additional physics in comparison to usual methods. In this work, an improved lattice Boltzmann model for phase transition flow will be introduced. First, the state of art for Shan and Chen (SC) type of LBM will be reminded. Then, link to real thermodynamics will be established with Maxwell equal areas construction. Convergence to isothermal liquid vapor equilibrium will be shown and discussed. Inclusion of an equation of state for real fluid and better incorporation of force term is presented. Multi-range interactions have been used for SC model, but it hasn't been yet applied to real fluid with non-ideal equation of state. In this work, we evaluate this model when it is applied to real liquid-vapor equilibrium. We show that important differences are found for evaluation of gas density. In order to recover thermodynamic consistency, we use a new scheme for calculation of force term, which is a combination of multi range model and numerical weighting used by Gong and Cheng. We show the superiority of our new model by studying convergence to equilibrium values over a large temperature range. We prove that spurious velocities remaining at equilibrium are decreased.
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
Characterization of the Atmosphere of Super-Earth 55 Cancri e Using High-resolution Ground-based Spectroscopy
The nearby super-Earth 55 Cnc e orbits a bright (V = 5.95 mag) star with a period of -18 hr and a mass of -8M {sub ⊕}. Its atmosphere may be water-rich and have a large-scale height; though, attempts to characterize it have yielded ambiguous results. Here we present a sensitive search for water and TiO in its atmosphere at high spectral resolution using the Gemini North telescope and the GRACES spectrograph. We combine observations with previous observations from Subaru and CFHT, improving the constraints on the presence of water vapor. We adopt parametric models with an updated planet radius based on recent measurements, and use a cross-correlation technique to maximize sensitivity. Our results are consistent with atmospheres that are cloudy or contain minimal amounts of water and TiO. Using these parametric models, we rule out a water-rich atmosphere (VMR ≥ 0.1%) with a mean molecular weight of ≤ 15 g mol{sup -1} at a 3σ confidence level, improving on the previous limit by a significant margin. For TiO, we rule out a mean molecular weight of ≤ 5 g mol{sup -1} with a 3σ confidence level for a VMR greater than 10{sup -8}; for a VMR of greater than 10{sup -7}, the limit rises to a mean molecular weight of ≤ 10 g mol{sup -1}. We can rule out low mean-molecular-weight chemical equilibrium models both including and excluding TiO/VO at very high confidence levels (>10σ). Overall, our results are consistent with an atmosphere with a high mean molecular weight and/or clouds, or no atmosphere.
79 ASTRONOMY AND ASTROPHYSICS↗
Di-CNN: Domain-Knowledge-Informed Convolutional Neural Network for Manufacturing Quality Prediction
In manufacturing, convolutional neural networks (CNNs) are widely used on image sensor data for data-driven process monitoring and quality prediction. However, as purely data-driven models, CNNs do not integrate physical measures or practical considerations into the model structure or training procedure. Consequently, CNNs’ prediction accuracy can be limited, and model outputs may be hard to interpret practically. This study aims to leverage manufacturing domain knowledge to improve the accuracy and interpretability of CNNs in quality prediction. A novel CNN model, named Di-CNN, was developed that learns from both design-stage information (such as working condition and operational mode) and real-time sensor data, and adaptively weighs these data sources during model training. It exploits domain knowledge to guide model training, thus improving prediction accuracy and model interpretability. A case study on resistance spot welding, a popular lightweight metal-joining process for automotive manufacturing, compared the performance of (1) a Di-CNN with adaptive weights (the proposed model), (2) a Di-CNN without adaptive weights, and (3) a conventional CNN. The quality prediction results were measured with the mean squared error (MSE) over sixfold cross-validation. Model (1) achieved a mean MSE of 6.8866 and a median MSE of 6.1916, Model (2) achieved 13.6171 and 13.1343, and Model (3) achieved 27.2935 and 25.6117, demonstrating the superior performance of the proposed model.
47 OTHER INSTRUMENTATION↗
Improving Climate Projections Using "Intelligent" Ensembles
Recent changes in the climate system have led to growing concern, especially in communities which are highly vulnerable to resource shortages and weather extremes. There is an urgent need for better climate information to develop solutions and strategies for adapting to a changing climate. Climate models provide excellent tools for studying the current state of climate and making future projections. However, these models are subject to biases created by structural uncertainties. Performance metrics-or the systematic determination of model biases-succinctly quantify aspects of climate model behavior. Efforts to standardize climate model experiments and collect simulation data-such as the Coupled Model Intercomparison Project (CMIP)-provide the means to directly compare and assess model performance. Performance metrics have been used to show that some models reproduce present-day climate better than others. Simulation data from multiple models are often used to add value to projections by creating a consensus projection from the model ensemble, in which each model is given an equal weight. It has been shown that the ensemble mean generally outperforms any single model. It is possible to use unequal weights to produce ensemble means, in which models are weighted based on performance (called "intelligent" ensembles). Can performance metrics be used to improve climate projections? Previous work introduced a framework for comparing the utility of model performance metrics, showing that the best metrics are related to the variance of top-of-atmosphere outgoing longwave radiation. These metrics improve present-day climate simulations of Earth's energy budget using the "intelligent" ensemble method. The current project identifies several approaches for testing whether performance metrics can be applied to future simulations to create "intelligent" ensemble-mean climate projections. It is shown that certain performance metrics test key climate processes in the models, and that these metrics can be used to evaluate model quality in both current and future climate states. This information will be used to produce new consensus projections and provide communities with improved climate projections for urgent decision-making.
Baker, Noel C.↗
Improving Climate Projections Using "Intelligent" Ensembles
Recent changes in the climate system have led to growing concern, especially in communities which are highly vulnerable to resource shortages and weather extremes. There is an urgent need for better climate information to develop solutions and strategies for adapting to a changing climate. Climate models provide excellent tools for studying the current state of climate and making future projections. However, these models are subject to biases created by structural uncertainties. Performance metrics-or the systematic determination of model biases-succinctly quantify aspects of climate model behavior. Efforts to standardize climate model experiments and collect simulation data-such as the Coupled Model Intercomparison Project (CMIP)-provide the means to directly compare and assess model performance. Performance metrics have been used to show that some models reproduce present-day climate better than others. Simulation data from multiple models are often used to add value to projections by creating a consensus projection from the model ensemble, in which each model is given an equal weight. It has been shown that the ensemble mean generally outperforms any single model. It is possible to use unequal weights to produce ensemble means, in which models are weighted based on performance (called "intelligent" ensembles). Can performance metrics be used to improve climate projections? Previous work introduced a framework for comparing the utility of model performance metrics, showing that the best metrics are related to the variance of top-of-atmosphere outgoing longwave radiation. These metrics improve present-day climate simulations of Earth's energy budget using the "intelligent" ensemble method. The current project identifies several approaches for testing whether performance metrics can be applied to future simulations to create "intelligent" ensemble-mean climate projections. It is shown that certain performance metrics test key climate processes in the models, and that these metrics can be used to evaluate model quality in both current and future climate states. This information will be used to produce new consensus projections and provide communities with improved climate projections for urgent decision-making.
Baker, Noel C.↗
Weighted Distances in Scale-Free Configuration Models
In this paper we study first-passage percolation in the configuration model with empirical degree distribution that follows a power-law with exponent τ∈(2,3). We assign independent and identically distributed (i.i.d.) weights to the edges of the graph. We investigate the weighted distance (the length of the shortest weighted path) between two uniformly chosen vertices, called typical distances. When the underlying age-dependent branching process approximating the local neighborhoods of vertices is found to produce infinitely many individuals in finite time—called explosive branching process—Baroni, Hofstad and the second author showed in Baroni et al. (J Appl Probab 54(1):146–164, 2017) that typical distances converge in distribution to a bounded random variable. The order of magnitude of typical distances remained open for the τ∈(2,3) case when the underlying branching process is not explosive. We close this gap by determining the first order of magnitude of typical distances in this regime for arbitrary, not necessary continuous edge-weight distributions that produce a non-explosive age-dependent branching process with infinite mean power-law offspring distributions. This sequence tends to infinity with the amount of vertices, and, by choosing an appropriate weight distribution, can be tuned to be any growing function that is O(loglog n), where n is the number of vertices in the graph. We show that the result remains valid for the the erased configuration model as well, where we delete loops and any second and further edges between two vertices.
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
Super-resolution and segmentation deep learning for breast cancer histopathology image analysis
Traditionally, a high-performance microscope with a large numerical aperture is required to acquire high-resolution images. However, the images’ size is typically tremendous. Therefore, they are not conveniently managed and transferred across a computer network or stored in a limited computer storage system. As a result, image compression is commonly used to reduce image size resulting in poor image resolution. Here, we demonstrate custom convolution neural networks (CNNs) for both super-resolution image enhancement from low-resolution images and characterization of both cells and nuclei from hematoxylin and eosin (H&E) stained breast cancer histopathological images by using a combination of generator and discriminator networks so-called super-resolution generative adversarial network-based on aggregated residual transformation (SRGAN-ResNeXt) to facilitate cancer diagnosis in low resource settings. The results provide high enhancement in image quality where the peak signal-to-noise ratio and structural similarity of our network results are over 30 dB and 0.93, respectively. The derived performance is superior to the results obtained from both the bicubic interpolation and the well-known SRGAN deep-learning methods. In addition, another custom CNN is used to perform image segmentation from the generated high-resolution breast cancer images derived with our model with an average Intersection over Union of 0.869 and an average dice similarity coefficient of 0.893 for the H&E image segmentation results. Finally, we propose the jointly trained SRGAN-ResNeXt and Inception U-net Models, which applied the weights from the individually trained SRGAN-ResNeXt and inception U-net models as the pre-trained weights for transfer learning. The jointly trained model’s results are progressively improved and promising. We anticipate these custom CNNs can help resolve the inaccessibility of advanced microscopes or whole slide imaging (WSI) systems to acquire high-resolution images from low-performance microscopes located in remote-constraint settings.
60 APPLIED LIFE SCIENCES↗
Steady induction effects in geomagnetism. Part 1C: Geomagnetic estimation of steady surficial core motions: Application to the definitive geomagnetic reference field models
In the source-free mantle/frozen-flux core magnetic earth model, the non-linear inverse steady motional induction problem was solved using the method presented in Part 1B. How that method was applied to estimate steady, broad-scale fluid velocity fields near the top of Earth's core that induce the secular change indicated by the Definitive Geomagnetic Reference Field (DGRF) models from 1945 to 1980 are described. Special attention is given to the derivation of weight matrices for the DGRF models because the weights determine the apparent significance of the residual secular change. The derived weight matrices also enable estimation of the secular change signal-to-noise ratio characterizing the DGRF models. Two types of weights were derived in 1987-88: radial field weights for fitting the evolution of the broad-scale portion of the radial geomagnetic field component at Earth's surface implied by the DGRF's, and general weights for fitting the evolution of the broad-scale portion of the scalar potential specified by these models. The difference is non-trivial because not all the geomagnetic data represented by the DGRF's constrain the radial field component. For radial field weights (or general weights), a quantitatively acceptable explication of broad-scale secular change relative to the 1980 Magsat epoch must account for 99.94271 percent (or 99.98784 percent) of the total weighted variance accumulated therein. Tolerable normalized root-mean-square weighted residuals of 2.394 percent (or 1.103 percent) are less than the 7 percent errors expected in the source-free mantle/frozen-flux core approximation.
Voorhies, Coerte V.↗
A transfer learning approach to energy-efficient control of small and medium-sized commercial buildings
Model-free reinforcement learning (RL) provides a data-driven and adaptive approach to optimize building energy use while satisfying occupant comfort. This powerful tool does not need any prior knowledge about the environment and system it is optimizing and can adapt its policy based on the changes in captures. Like any other data-driven tool, it faces high training costs due to the extensive agent-environment interactions required to capture long-term building dynamics and user comfort. Transfer learning, particularly policy distillation, offers a promising way to accelerate training by leveraging pretrained RL agents in different building and system types. Here, this study investigates online student distillation, in which the student model updates its neural network weights using outputs from teacher models. The work introduces a student distillation strategy designed for efficient knowledge transfer, along with a teacher selection method that ensures high-quality guidance. The approach is validated using a highly calibrated whole building energy model for a small/medium commercial building test facility. Results show substantial reductions in training time and data requirements while surpassing the performance of ASHRAE Guideline 36, an advanced rule-based control strategy. The distilled RL model required 45% less data and achieved 20% higher cumulative rewards than a state-of-the-art RL model, with faster convergence and lower energy consumption. These outcomes demonstrate that effective transfer learning enables a scalable and data-efficient energy management solution for commercial buildings.
ASHRAE guideline 36↗
Real-time Adaptive Control Using Neural Generalized Predictive Control
The objective of this paper is to demonstrate the feasibility of a Nonlinear Generalized Predictive Control algorithm by showing real-time adaptive control on a plant with relatively fast time-constants. Generalized Predictive Control has classically been used in process control where linear control laws were formulated for plants with relatively slow time-constants. The plant of interest for this paper is a magnetic levitation device that is nonlinear and open-loop unstable. In this application, the reference model of the plant is a neural network that has an embedded nominal linear model in the network weights. The control based on the linear model provides initial stability at the beginning of network training. In using a neural network the control laws are nonlinear and online adaptation of the model is possible to capture unmodeled or time-varying dynamics. Newton-Raphson is the minimization algorithm. Newton-Raphson requires the calculation of the Hessian, but even with this computational expense the low iteration rate make this a viable algorithm for real-time control.
Haley, Pam↗
SEE Rate Estimation: Model Complexity and Data Requirements
Statistical Methods outlined in [Ladbury, TNS20071 can be generalized for Monte Carlo Rate Calculation Methods Two Monte Carlo Approaches: a) Rate based on vendor-supplied (or reverse-engineered) model SEE testing and statistical analysis performed to validate model; b) Rate calculated based on model fit to SEE data Statistical analysis very similar to case for CREME96. Information Theory allows simultaneous consideration of multiple models with different complexities: a) Model with lowest AIC usually has greatest predictive power; b) Model averaging using AIC weights may give better performance if several models have similar good performance; and c) Rates can be bounded for a given confidence level over multiple models, as well as over the parameter space of a model.
Ladbury, Ray↗
The Impacts of Rising Temperatures on Aircraft Takeoff Performance
Steadily rising mean and extreme temperatures as a result of climate change will likely impact the air transportation system over the coming decades. As air temperatures rise at constant pressure, air density declines, resulting in less lift generation by an aircraft wing at a given airspeed and potentially imposing a weight restriction on departing aircraft. This study presents a general model to project future weight restrictions across a fleet of aircraft with different takeoff weights operating at a variety of airports. We construct performance models for five common commercial aircraft and 19 major airports around the world and use projections of daily temperatures from the CMIP5 model suite under the RCP 4.5 and RCP 8.5 emissions scenarios to calculate required hourly weight restriction. We find that on average, 10 - 30% of annual flights departing at the time of daily maximum temperature may require some weight restriction below their maximum takeoff weights, with mean restrictions ranging from 0.5 to 4% of total aircraft payload and fuel capacity by mid- to late century. Both mid-sized and large aircraft are affected, and airports with short runways and high temperatures, or those at high elevations, will see the largest impacts. Our results suggest that weight restriction may impose a non-trivial cost on airlines and impact aviation operations around the world and that adaptation may be required in aircraft design, airline schedules, and/or runway lengths.
air transportation system↗
Machine Learning for Memory Reduction in the Implicit Monte Carlo Simulations of Thermal Radiative Transfer [Slides]
Project Goal: Use parametric Machine Learning methods in order to reduce memory requirements at checkpointing & restarting in the IMC simulations of Thermal Radiative Transfer using: Expectation Maximization and Weighted Gaussian Mixture Model-based approach for `particle-data compression', introduced in Plasma Physics to model Maxwellian particle distributions by Luis Chacon and Guangye Chen; Expectation Maximization with Weighted Hyper-Erlang Model in order to compress isotropic IMC particle data in the frequency domain; and Expectation Maximization and von Mises-Fisher Mixture Model for compression of anisotropic IMC particle data in the angular domain (work-in-progress).
73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
The use of a block diagram simulation language for rapid model prototyping
The research performed this summer focussed on the development of a predictive model for the loading of liquid oxygen (LO2) into the external tank (ET) of the shuttle prior to launch. A predictive model can greatly aid the operational personnel since instrumentation aboard the orbiter and ET is limited due to weight constraints. The model, which focuses primarily on the orbiter section of the system was developed using a block diagram based simulation language known as VisSim. Simulations were run on LO2 loading data for shuttle flights STS50 and STS55 and the model was demonstrated to accurately predict the sensor data recorded for these flights. As a consequence of the simulation results, it can be concluded that the software tool can be very useful for rapid prototyping of complex models.
Whitlow, Jonathan E.↗