Session MB: Turbulence Control II

Characterization of flow control effects over a 3D turret

Separated flow phenomena around a turret causes optical distortions on a laser beam. In an effort to minimize these effects, active flow control has been applied. On this effort, experiments ran at Syracuse University wind tunnel included tests at Reynolds number of 300,000, based on diameter and free-stream velocity, using a cylindrical turret. The model has 6 inches in diameter with a flat aperture of 2.8in diameter. The actuation system consists of 11 synthetic jets placed fore of a flat aperture facing the incoming flow. Particle Image Velocimetry was used to capture the 2-component velocity field over the turret. Simultaneous surface pressure and velocity measurements were acquired for flow characterization. Previous velocity results have shown a positive effect of the actuation on reducing the turbulence intensity observed at the center plane. Integral length scales also present the actuation effect in a periodic pattern observed that seems to have the actuation frequency. A reduction on the integral scales is also observed.   

The Effect of Closed Loop Flow Control on the Integral Scales over a 3D Turret

An extended active flow control study was conducted with a 3D turret with a flat aperture in the Subsonic Aerodynamic Research Laboratory (SARL) windtunnel at Wright-Patterson Air Force Base. The pressure based integral scales of the flow over the aperture are altered by using a simple, pressure based proportional closed loop controller. Control authority was obtained with actuators mounted fore of the aperture while pressure transducers mounted on and aft of the aperture provided observability of the system. For closed loop controller, the bandpass filtered temporal POD mode coefficients of the surface pressure were used as the feedback signal. Multiple control cases were examined using various POD methods such as the baseline, lumped, and split POD. Comparing the flow case with no actuation with the flow cases with the various closed loop control, a clear reduction of the integral scales is seen in each case.   

Application of closed-loop control techniques in an axisymmetric jet at Mach 0.6

One of the more challenging and far-reaching aspects of fluid dynamics remains the prediction and ultimate control of highly turbulent, non-linear, flow physics. As is pertains to acoustic-noise, the challenge arises in pinpointing the mechanisms within these high-speed flows which are the most efficient propagators of highly intense acoustic pressure fluctuations that translate to the far-field as broadband noise. For the experiment, a simple pressure based proportional closed-loop feedback controller was implemented to manipulate the shear layer of the jet flow. Previous work by Hall \textit{et al.} demonstrated that the modal characteristics of the near-field pressure, sampled within the noise producing region ($x/D=6-10)$, were uniquely correlated with the far-field acoustics. The helical azimuthal mode (\textit{mode-1}) diminished the strength of the correlated signature. The pressure is Fourier transformed and the azimuthal mode-1 signature is fed back, using a simple proportional closed-loop feedback controller. This experimental setup makes it possible to manipulate the shear layer of the flow field with a forcing pattern of azimuthal mode-1, as well as directly utilizing the amplitude and frequency characteristics of mode-1 perturbations from the sampled pressure field.   

A new passive turbulence grid with improved isotropy

To experimentally produce isotropic and near-homogeneous turbulence, a multitude of grids of various complexity have been used. The best results so far have been obtained with the most complex active grids. Here we propose yet another type of grid: a \emph{passive} grid with tethered spheres attached at each mesh corner. The simplicity of this new grid is particularly interesting in facilities where active grids and secondary contractions after the grid cannot be implemented. Statistical quantities for different configurations are measured with a two-component LDA and compared to the plain grid without attachments. It is shown that the tethered spheres not only improve the balance of the average kinetic energies of the longitudinal and the transverse velocity components $u^2_{rms}$ and $v^2_{rms}$ but also the spectral isotropy $E_{uu}/E_{vv}$.   

A new approach to: (a) grid generation for numerical optimization, and (b) interconnect networks for beowulf clusters, leveraging n-dimensional sphere-packings

The abstract field of n-dimensional sphere packing theory is well developed (for a comprehensive review, see Sphere Packings, Lattices and Groups by Conway and Sloane). This theory forms the theoretical underpinning of the error-correcting codes used in both deep space communications and in computer memory. The present work extends this elegant theory to two important and immensely practical problems in computational fluid dynamics: (a) the generation of efficient grids for the coordination of grid-based derivative-free optimization algorithms in n dimensions, and (b) the effective n-dimensional interconnection of massively-parallel clusters of computational nodes. As we will illustrate and quantify, the first problem benefits tremendously from dense sphere packings with large kissing numbers $>>$ 2n, whereas the latter problem benefits tremendously from rare sphere packings with kissing number = n+1.   

A lattice-based approach to derivative-free optimization

The optimization of an expensive, high-dimensional function when no derivative information is available necessitates the use of a derivative-free optimization algorithm. Such a scenario is evident, for example, when optimizing a finite-time-average approximation of an infinite-time-average statistic of a chaotic system such as a turbulent flow. The truncation error induced by such an approximation renders the calculation of the derivative ineffective. Due to the often significant expense associated with performing repeated function evaluations, a derivative-free optimization algorithm which converges to within an accurate tolerance of the global minimum of a nonconvex function of interest with a minimum number of function evaluations is desired. One of the most efficient algorithms available, known as the Surrogate Management Framework, combines a grid-based pattern search Poll step with inexpensive interpolating ``surrogate'' functions to provide suggested regions of parameter space in which to perform new function evaluations. The present work considers an SMF algorithm that combines a pattern search based on N-dimensional sphere packings, or lattices, with a highly efficient surrogate search. The lattice-based Poll step offers substantially greater efficiency compared to previous Cartesian grid-based algorithms; combined with an extremely effective Search, a unique, highly efficient SMF algorithm has been devised.   

Ensemble/Variational Estimation (EnVE) and its application to turbulent flows in complex geometries

A new algorithm, Ensemble/Variational Estimation (EnVE), has been developed as a consistent hybrid data assimilation method that combines the nonlinear statistical propagation properties of the Ensemble Kalman Filter (EnKF) and the retrospective analysis capabilities of 4DVar/Moving Horizon Estimation (MHE). A sophisticated C++ object-oriented framework has been developed that implements the EnVE algorithm to facilitate its application to any complex (multiscale/multiphysics) flow code of interest in a highly parallel fashion with minimal changes to the existing flow solver. In the present work, this framework has been applied to the flagship unstructured LES code (CDP) developed at the Center for Integrated Turbulence Simulations (CITS) at Stanford University.   

Ensemble/Variational Estimation (EnVE) and its application to canonical turbulent flow realizations

The recently-developed hybrid EnVE method for data assimilation incorporates successive adjoint optimizations to update the initial conditions of a flow model, over various horizons of interest, in order to reconcile this model with recent measurements. Such adjoint optimizations typically require the trajectory to be saved over the entire interval over which the optimization is performed; in high-dimensional systems, this can lead to significant storage problems, which can be partially alleviated via checkpointing. In the EnVE framework, this requirement is eliminated, and supplanted by a requirement to march the state of the system backward in time simultaneously with the adjoint. If the system is derived from a PDE with a diffusive component, this backward-in-time state march is ill conditioned, and requires regularization/smoothing to prevent errors from accumulating rapidly at the small scales. The present talk focuses on this peculiar requirement of the EnVE algorithm. As the forecasting problem may itself be considered as a smoothing problem, it is, in fact, expected to find a ``smoothing'' ingredient at the heart of an algorithm of this sort. Various strategies are proposed and tested for accomplishing the required smoothing in the EnVE setting, and are tested on both a chaotic 1D PDE (the Kuramoto-Sivashinsky equation) as well as our in-house spectral 3D DNS/LES code, diablo.   

Ensemble/Variational Observation (EnVO): a rigorous approach to adaptive observation built on the framework of EnVE

Building directly on the foundation of our group's hybrid Ensemble/Variational Estimation (EnVE) algorithm, a new algorithm is proposed for adaptive observation. This algorithm, EnVO, is based on the application of adjoint-based model predictive control (MPC) applied to an ensemble of realizations in order to minimize a well targetted objective: specifically, a quantification of the estimate uncertainty at the forecast time. This estimate uncertainty, in turn, is quantified using the EnKF procedure by marching the ensemble forward to the forecast time while assimilating the anticipated measurements along the candidate sensor trajectories. A noncooperative optimization framework is also outlined in order to robustify this optimization. We show how this algorithm may be applied to optimize the trajectories of several sensor-equipped unmanned aerial vehicles (UAVs) in order to measure and forecast an atmospheric contaminant release plume in a maximally effective manner.   

Pseudo Volume-filling Sampling (PVS) via Bouyancy Control in Hurricane and Ocean Systems

This project addresses improved sensing methods for modeling and prediction in two seemingly disparate applications, hurricane systems and ocean currents. In such applications, well distributed measurements of important flow quantities such as the local fluid velocity, temperature, and pressure are quite valuable. Such measurements can be obtained from very simple sensor systems. For example, thousands of sensor-equipped balloons may be released by a cargo plane making a single pass over a hurricane. Thousands of sensor-equipped floats have in fact already been distributed over the oceans of the earth in the Argo project. The challenge in both cases is to distribute the sensors uniformly over the volumes of interest. In both cases, the sensor systems are grossly underactuated, and are only capable of controlling their vertical motion. Given that an accurate estimate of the background velocity field is available in both applications, however, it is possible to ``fly'' each individual sensor system, much as a recreational balloonist can direct a balloon accurately by exploiting known velocity shear within the atmosphere. The present work addresses how an entire network of such underactuated sensor systems can distributed uniformly over a domain of interest, using both global model predictive control (MPC) regulated centrally and, in certain well-defined subproblems, simple LQG-based strategies implemented locally.