SEARCH · Search NASA
Results for “predictive model”
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.
Real-Time Model Predictive Powertrain Control for a Connected Plug-In Hybrid Electric Vehicle
Not provided.
hIPPYlib-MUQ: A Bayesian Inference Software Framework for Integration of Data with Complex Predictive Models under Uncertainty
Bayesian inference provides a systematic framework for integration of data with mathematical models to quantify the uncertainty in the solution of the inverse problem. However, the solution of Bayesian inverse problems governed by complex forward models described by partial differential equations (PDEs) remains prohibitive with black-box Markov chain Monte Carlo (MCMC) methods. We present hIPPYlib-MUQ, an extensible and scalable software framework that contains implementations of state-of-the art algorithms aimed to overcome the challenges of high-dimensional, PDE-constrained Bayesian inverse problems. These algorithms accelerate MCMC sampling by exploiting the geometry and intrinsic low-dimensionality of parameter space via derivative information and low rank approximation. The software integrates two complementary open-source software packages, hIPPYlib and MUQ. hIPPYlib solves PDE-constrained inverse problems using automatically-generated adjoint-based derivatives, but it lacks full Bayesian capabilities. MUQ provides a spectrum of powerful Bayesian inversion models and algorithms, but expects forward models to come equipped with gradients and Hessians to permit large-scale solution. By combining these two complementary libraries, we created a robust, scalable, and efficient software framework that realizes the benefits of each and allows us to tackle complex large-scale Bayesian inverse problems across a broad spectrum of scientific and engineering disciplines. To illustrate the capabilities of hIPPYlib-MUQ, we present a comparison of a number of MCMC methods available in the integrated software on several high-dimensional Bayesian inverse problems. These include problems characterized by both linear and nonlinear PDEs, various noise models, and different parameter dimensions. The results demonstrate that large (~ 50×) speedups over conventional black box and gradient-based MCMC algorithms can be obtained by exploiting Hessian information (from the log-posterior), underscoring the power of the integrated hIPPYlib-MUQ framework.
Hot, Cold, or Just Right? Model Predictive Controller (HCJR MPC) v0.1.0
The project seeks to develop new sensor and control technologies that will improve occupant thermal comfort and reduce cooling energy consumption in commercial buildings by minimizing overcooling.
Overland flow numerical model prediction, Lower Triangle Region in East River Watershed, Colorado, 3 days
This data package contains numerical simulation results of surface flow variables such as flow velocity and water depth in Lower Triangle Region in East River Watershed, Colorado. The surface flow is a consequence of a high intensity rainfall event with a total duration of 3 days, available at a resolution of 10 minutes. The results are computed on triangular multiresolution meshes with resolutions ranging from 10 meter to 80 meter. The data package also contains a simulation on a uniform triangular mesh with a resolution of 10 meter. The simulations consider surface flow only and neglect subsurface flow, infiltration, and evapotranspiration. The purpose of the data is to assess the quality of a mesh refinement strategy.
Predictive Modeling of Polymer Mechanical Behavior Coupled to Chemical Change/ Technique Development for Measuring Polymer Physical Aging.
Develop an understanding of the evolution of glassy polymer mechanical response during aging and the mechanisms associated with that evolution. That understanding will be used to develop constitutive models to assess the impact of stress evolution in encapsulants on NW designs.
Predictive Modeling, Inverse Problems, and Uncertainty Quantification with Application to Emergency Response
No abstract provided.
Predictive Modeling of CO 2 Sequestration and Storage in Deep Saline Sandstone Reservoirs: Sensitivity Analysis of Mineral Rates in Reactive Transport
One idea for mitigating the increase in fossil- fuel generated carbon dioxide (CO 2 ) in the atmosphere is to inject CO 2 into subsurface saline sandstone reservoirs, thereby storing it in those geologic formations and out of the atmosphere.
The use of atmospheric prediction models to invert infrasound for linear-equivalent time domain moment tensors: Source Physics Experiment, Phase 1
We invert far field infrasound data for the equivalent seismo-acoustic time domain moment tensor to assess the effects of variable atmospheric models as well as to quantify the relative contributions of two presumed source phenomena. The infrasound data was produced by a series of underground chemical explosions that were conducted during the Source Physics Experiment, (SPE) which was originally designed to study explosion-generated seismo-acoustic signal phenomena. The goal of the work presented herein is two-fold: the first goal is to investigate the sensitivity of the estimated time domain moment tensors to variability of the estimated atmospheric model. The second goal is to determine the relative contribution of two possible source mechanisms to the observed infrasonic wave field. Rather than using actual atmospheric observations to estimate the necessary atmospheric Green's functions, we build a series of atmospheric models that rely on publicly available, regional atmospheric observations and the assumption that the acoustic energy results from a linear combination of an underground isotropic explosion and surface spall. The atmospheric observations are summarized and interpolated onto a 3D grid to produce a model of sound speed at the time of the experiment. For each of four SPE acoustic datasets that we invert, we produced a suite of three atmospheric models, based on ten years of regional meteorological observations: an average model, which averages the atmospheric conditions for ten years prior to each SPE event, as well as two extrema models. We find that the inversion yields relatively repeatable results for the estimated spall source. Conversely, the estimated isotropic explosion source is highly variable. This suggests that the majority of the observed acoustic energy is produced by the spall source and/or our modeling of the elastic energy propagation, and it's subsequent conversion to acoustic energy via linear elastic-to-acoustic coupling at the free surface, is too simplistic.
PakMAN: A Predictive Model of Political Instability in Pakistan [Slides]
Abstract not provided.
Predictive Models and High-Performance Computing as Tools to Accelerate the Scaling-Up of New Bio-Based Fuels Workshop: Summary Report
This report summarizes the results of a virtual workshop sponsored by the Bioenergy Technologies Office held on June 9–11, 2020. The workshop discussed best practices for utilizing mathematical modeling tools across multiple scales to reduce technology uncertainty and accelerate scaling-up of biorefinery/chemical production equipment and optimize operations.
Development of Predictive Models for Fleet Reliability in the Aircraft Industry
Population and sample databases were created with FAA Airworthiness Directives (AD) from 2000 to 2017 to identify failure points in the aircraft industry. Three linear models were created using these databases to find trends that could inform decision makers about the general nature of fleet recalls. The first relates the percent of the fleet affected to the subsystem listed in the AD, the second relates the severity of damage to the aircraft with respect to the subsystem of the aircraft, and the third relates the age of the aircraft to the time of the AD. These models concluded that younger aircraft experience more severe recalls, for all aircraft the severity of damage decreased as more of the fleet was affected by an AD, and active systems experience more severe ADs than passive systems. These trends show that there are significant relationships that exist that can be used in other areas of design.
(U) A Linear Response Model Predicts Reactivity From a Density Profile
We tested the ability to predict the system reactivity, described by alpha, given a density profile using a simple linear system response. We generated a suite of 1-dimensional density profiles that consisted of nominal density, a discontinuity, and a decay. These profiles were prescribed a functional form and the mass was conserved in all cases. From these density profiles, we calculated the alpha value of the 3-dimensional system.We calculated a linear response function given a training set of the 1-dimensional density profiles, and the system reactivity described by alpha. We tested the robustness of the response function using the remaining test data. Our results showed very good agreement between the predicted and calculated test values, where the distribution of alpha differences was centered about zero and had a standard deviation of 0.005 gens/shake. The predicted and calculated alpha values did not significantly differ (t=-0.0009 p<0.99). We used Singular Value Decomposition (SVD) to reduce the matrix rank by retaining95% of the cumulative singular value contributions. This reduced the matrix rank by 91.7%. We generated the linear response matrix and calculated the difference between the predicted and calculated alpha values. Using the reduced order matrix, we showed good agreement between the predicted and calculated alpha values where the distribution of differences was centered near zero, the standard deviation was 0.006 gens/shake, and the statistical t-test showed good agreement (t=0.02, p<0.98). These results show a linear relationship between a series of 1-dimensional density profiles,where the mass was conserved, and the system reactivity. The next steps of this work will be to investigate the linear response using 2-dimensional density profiles.
Mapping hydrologic and biogeochemical information flows to improve predictive models and understand climate influence
Focal area: Insights gleaned from complex data (both observed and simulated) using AI, big data analytics, and other advanced methods, including explainable AI and physics or knowledge-guide AI.
Predictive Models and Novel Accelerated Tests for the Reliability of Cell Metallization and Solder Joints in Photovoltaic Modules
Interconnect, solder, and metallization-related failures are the primary reason for premature Si-based photovoltaic panel failure [1, 2] and can lead to a cascade of other failures, including thermal events. This combination of high severity and high occurrence means that understanding interconnect-related failures is key to the long-term health of the photovoltaic industry, and it is crucial that the industry adopt efficient and effective tests that aid in design-for-reliability and manufacturing quality control.
Model Predictive Integrated Voltage and Frequency Support in Microgrids.
Abstract not provided.