Browsing by Subject "uncertainty quantification"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Development of Interatomic Potentials with Uncertainty Quantification: Applications to Two-dimensional Materials
    (2019-07) Wen, Mingjian
    Atomistic simulation is a powerful computational tool to investigate materials on the microscopic scale and is widely employed to study a large variety of problems in science and engineering. Empirical interatomic potentials have proven to be an indis- pensable part of atomistic simulation due to their unrivaled computational efficiency in describing the interactions between atoms, which produce the forces governing atomic motion and deformation. Atomistic simulation with interatomic potentials, however, has historically been viewed as a tool limited to provide only qualitative insight. A key reason is that in such simulations there are many sources of uncertainty that are difficult to quantify, thus failing to give confidence interval on the obtained results. This thesis presents my research work on the development of interatomic potentials with the ability to quantify the uncertainty in simulation results. The methods to train interatomic po- tentials and quantify the uncertainty are demonstrated via two-dimensional materials and heterostructures throughout this thesis, whose low-dimensional nature makes them distinct from their three-dimensional counterparts in many aspects. Both physics-based and machine learning interatomic potentials are developed for MoS2 and multilayer graphene structures. The new potentials accurately model the interactions in these systems, reproducing a number of structural, energetic, elastic, and thermal properties obtained from first-principles calculations and experiments. For physics-based poten- tials, a method based on Fisher information theory is used to analyze the parametric sensitivity and the uncertainty in material properties obtained from phase average. We show that the dropout technique can be applied to train neural network potentials and demonstrate how to obtain the predictions and the associated uncertainties of material properties practically and efficiently from such potentials. Putting all these ingredients of my research work together, we create an open-source fitting framework to train inter- atomic potentials and hope it can make the development and deployment of interatomic potentials easier and less error prone for other researchers.
  • Thumbnail Image
    Item
    Quantification and reduction of uncertainty of model predictions of wind turbines and plants via high-fidelity simulations
    (2016-12) Foti, Daniel
    With increasing energy demands renewable energy sources are continuing to receive attention and investment to become a larger source for electricity production. Today, wind generated power through wind turbines creates 4% of the electricity in the United States. The wind energy share of the electricity market is expected to grow rapidly as the United States Department of Energy goal is to reach 20% wind generated electricity by 2030. Computational models for wind plants can be used to predict wind plant performance and optimize the turbine placement and controls. However, uncertainties associated with such models, due to, among others, the computationally expedient simplifications need to be carefully assessed, quantified and reduced. A numerical investigation of model wind turbines employing large-eddy simulation and the curvilinear immersed boundary method to resolve the geometrical details of the turbine is undertaken revealing that the unstable hub vortex interacts with the turbine tip shear layer. Using a spatio-temporal filtering technique, wake meandering, a large scale displacement of the wake, is reconstructed into three-dimensional helical meander profiles. Statistics of the amplitudes and wavelengths corresponding to the intensity and streamwise elongation of the periodic wake meandering indicate complex coherent structures. Similar simulations are performed using the computationally expedient wind turbine actuator surface models with and without a nacelle model to parameterize the turbine. All simulations are validated against substantial experimental measurements. The simulations with the nacelle model are able to accurately capture the geometry dependent near wake and the dynamics in the far wake. The simulations without the nacelle model predict a stable, columnar hub vortex which does not interact with the turbine tip shear layer. Moreover, the amplitude of the meandering profiles is shown to be larger in the immersed boundary method simulations and simulations with a nacelle model compared to the simulation without the nacelle model proving that the nacelle and unstable hub vortex augment the meandering intensity in wind turbines. Due to the exceptional performance of the computationally efficient actuator surface with nacelle model, several turbine designs are simulated with diameters ranging from the laboratory scale (0.1 meters) to the utility scale (96 meters). Despite significant geometrical differences, a characteristic velocity based on the turbine thrust collapses the profile of both the wake turbulence kinetic energy and the amplitude of wake meandering based on the meandering profile for all turbine sizes. This result suggests that the turbulence levels and wake meandering intensity are explicitly linked. The wavelengths of wake meandering are properly scaled by the diameter of the turbine. In agreement with numerous measurements, the wake meandering and hub vortex Strouhal number based on the incoming hub height velocity and diameter is found to be approximately 0.3 and 0.7, respectively, for all turbines. Dynamic mode decomposition of the velocity field indicates that the modes related to these frequencies contain a majority of the energy in the meandering wake and confirms that an unstable hub vortex is a necessary requirement for simulating wind turbine wakes. The Horn Rev offshore wind plant is investigated showing conclusive evidence that the nacelle and hub vortex are important in large arrays of wind turbines. The consistency across scales and wind plant rows of the stochastic distributions of the wake meandering amplitudes and wavelengths allows for the development of a reduced-order kinematic wake model with statistic-based wake meandering inputs. Finally, uncertainties in the model parameters or model inadequacy are investigated using a framework of non-intrusive polynomial chaos. The feasibility of using a kinematic wake model is determined by investigating the parameter uncertainty of surface roughness and induction factor. The parameter uncertainty of the nacelle model is considered in a series of large-eddy simulations. The aleatoric uncertainty of the surface friction on the model and the epistemic nacelle geometry uncertainty propagate downstream in the inner wake and have implications on the uncertainty of the turbulence levels in the entire far wake.