Proper Orthogonal Decomposition Results Research Articles

Closed-loop plasma flow control of a turbulent cylinder wake flow using machine learning at Reynolds number of 28 000

Machine learning is increasingly used for active flow control. In this experimental study, alternating-current dielectric barrier discharge plasma actuators are deployed for the closed-loop intelligent control of the flow around a cylinder at a Reynolds number of 28 000 based on the velocity feedback from two hot-wire sensors placed in the wake. Variations in the cylinder drag are monitored by a load cell, and the temporal response of the wake flow field is visualized by a high-speed particle image velocimetry system working at 1 kHz. The high-speed control law is operated using a field programmable gate array optimized by genetic programing (GP). The results show that the peak drag reduction achieved by machine learning is of similar magnitude to that of conventional steady actuation (∼15%), while the power saving ratio is 35% higher than with conventional techniques because of the reduced power consumption. Analysis of the best GP control laws shows that the intensity of plasma actuation should be kept at a medium level to maximize the power-saving ratio. When compared with the baseline uncontrolled flow, the best controlled cases constrain the meandering motion of the cylinder wake, resulting in a narrow stabilized velocity deficit zone in the time-averaged sense. According to the results of proper orthogonal decomposition and dynamic mode decomposition, Karman vortex shedding is promoted under the best GP control.

Supersonic cavity shear layer control using spanwise pulsed spark discharge array

An experimental study on supersonic cavity flow control using a spanwise pulsed spark discharge array (SP-PSDA) is performed in this paper. High-speed schlieren imaging at a frame rate of 50 kHz is deployed for flow visualization. The schlieren snapshots, as well as their statistics, are analyzed to reveal the supersonic cavity flow control effect and its underlying mechanism. Results show that the shear layer presents a wave-like oscillation due to thermal bulbs induced by SP-PSDA. Specifically, the shear layer structure in the baseline case resembles an incomplete hairpin structure, which becomes complete after plasma actuation. SP-PSDA actuation at 5 kHz has a better control effect, which enhances the IRMS of the whole hairpin structure and produces several channels within it—these aid momentum transport within the shear layer. According to the results of proper orthogonal decomposition, the thermal bulbs couple with the shear layer to form large-scale coherent structures. These structures excite the Kelvin–Helmholtz instability, converting the oscillation frequency of the shear layer to an actuation frequency.

Active Flow Control Based Bleed in an Axial Compressor Cascade Using Large Eddy Simulation

The performance of the compressor is of great importance in developing advanced engine technology, and it dramatically influences the cycle efficiency and safety of the engine. The compressor bleed requires a part of the main flow extraction from the passage, and the removal of the airflow plays an essential role in the performance of the compressor. The influence of the constant suction and pulsed suction on the flow structures and the aerodynamic performance in a compressor cascade, and the control mechanisms are numerically studied using large eddy simulation. At a suction mass flow rate of 0.25% of the main flow rate, constant suction leads to a 9.36% reduction in Cpt, while pulsed suction gains 24% more benefit in Cpt reduction on the basis of the constant suction with an excitation frequency of 100 Hz. It indicates that the pulsed suction has more significant potential to control the loss generation in the compressor. Afterward, the flow field of the baseline, constant suction, and pulsed suction are compared to find out the flow control mechanisms of the constant suction and pulsed suction. Two different formation mechanisms of the passage vortex are found due to the excitation effect of the pulsed suction. Besides, the results of proper orthogonal decomposition and dynamic mode decomposition show that the pulsed suction can simplify and stabilize the compressor flow field, which could lead to a further benefit in Cpt.

Clustering sparse sensor placement identification and deep learning based forecasting for wind turbine wakes

A data-driven approach is an alternative to extract general models for wind energy applications. A spatial sensitivity analysis is achieved using a probabilistic model to quantitatively identify the variability in performance due to individual parameters and visualize spatial distributions. Proper orthogonal decomposition results are combined with linear discriminant analysis under the clustering framework to present low-dimensional classifiers. Using the decomposition enables the system to be far away from ill-conditioned states. The optimal sensor locations are explicitly distributed in the transition region, where the velocity and Reynolds stresses relax toward a wake recovered state. With the optimal sensors, the cluster assignment and flow dynamics are obtained. There is an advantage in including more features in the reconstruction process to capture the slow and fast dynamics. Assessing the differences in the wake response and establishing the importance of spatial sensitivities are provided here for seeking accurate models. The bidirectional neural network is used to predict the fluctuating velocity of the considered sensors. The result of long–short term memory shows correlations of 92% between the real and predicted fluctuating velocities.

Investigation on aerodynamic characteristics of tailing vehicle hood in a two-vehicle platoon

Experimental and numerical methods were performed to study the hood aerodynamics of a 1:18 scale vehicle model in a two-vehicle platoon under different fixed intra spacing and dynamic intra spacing at the Reynolds number of 4.9 × 105 and 4.08 × 105 in a 1:15 scale wind tunnel. Blockage ratio was calculated to be 5.6% for experiments. The averaged and fluctuating pressures of the hood of the tailing vehicle are much higher than that of single vehicle, and stronger fluctuation occurs at the front and rear edges of hood. Pressure fluctuation over the tailing vehicle hood increases as the intra spacing diminishes but sharply declines within intra spacing of 0.2 L. The energy of fluctuation is concentrated in frequency bands from 50 to 110 Hz at wind speed of 30 m/s and 40–90 Hz at wind speed of 25 m/s. Energy at the main frequency bands and peak frequencies take up nearly 25% of total energy, respectively. The results of proper orthogonal decomposition show that total energy proportion is 49.6% for the first 11 modes and 67.0% for the first 30 modes, respectively. Relative motion at low speed has no discernible effect on hood behavior and peak frequencies have inclination to decline during the process of vehicle approaching.

On the streamwise and normal vortices forming on plates with spanwise periodic cambering

Three-dimensional measurements of the flow field over rigid plates with periodic cambering in the spanwise direction were performed. Two plates of an aspect ratio of 5 with 5 and 9 periodic cells were used, and measurements were performed at a Reynolds number of 28,000 and three angles of attack. Previous research showed that organized vortices form in the valleys between the periodic cells, and that these vortices contribute to slight enhancement of lift and reduction in drag. In this investigation, the details of these vortices are examined and compared at three different flow fields including: 1) short separation bubble, 2) long separation bubble and 3) fully separated flow. The formation of these primary vortices is observed in all cases. The formation of these vortices is influenced by the interaction between the periodic cambering, and the thickness and strength of the backflow. This interaction becomes strongest when a long separation bubble forms resulting in larger vortices in size and magnitude. The instantaneous flow field shows that the primary vortices form most of the time, while they move in the spanwise direction and breakdown in the streamwise direction at other instances of time. In addition to the primary vortices, smaller vortices also form in the vicinity of the primary vortices. Proper Orthogonal Decomposition results indicate that reconstructing the flow field with different number of modes isolates the primary vortices from the smaller scale vortices.

Analysis of Self-Excited Combustion Instabilities Using Decomposition Techniques

Proper orthogonal decomposition and dynamic mode decomposition are evaluated for the study of self-excited longitudinal combustion instabilities in laboratory-scaled single-element gas turbine and rocket combustors. Since each proper orthogonal decomposition mode comprises multiple frequencies, specific modes of the pressure and heat release are not related, which makes the analysis more qualitative and less efficient for identifying physical mechanisms. On the other hand, dynamic mode decomposition analysis generates a global frequency spectrum in which each mode corresponds to a specific discrete frequency so that different dynamics can be correlated. In addition, proper orthogonal decomposition results are found to be inaccurate when only a limited amount of spatial information is provided in contrast with dynamic mode decomposition results, which provide more reliable results. Overall, dynamic mode decomposition analysis proves to be a robust and systematic method that can give consistent interpretations of the periodic physics underlying combustion instabilities.

Chaotic vibrations of a post-buckled L-shaped beam with an axial constraint

Analytical results are presented on chaotic vibrations of a post-buckled L-shaped beam with an axial constraint. The L-shaped beam is composed of two beams which are a horizontal beam and a vertical beam. The two beams are firmly connected with a right angle at each end. The beams joint with the right angle is attached to a linear spring. The other ends are firmly clamped for displacement. The L-shaped beam is compressed horizontally via the spring at the beams joint. The L-shaped beam deforms to a post-buckled configuration. Boundary conditions are required with geometrical continuity of displacements and dynamical equilibrium with axial force, bending moment, and share force, respectively. In the analysis, the mode shape function proposed by the senior author is introduced. The coefficients of the mode shape function are fixed to satisfy boundary conditions of displacements and linearized equilibrium conditions of force and moment. Assuming responses of the beam with the sum of the mode shape function, then applying the modified Galerkin procedure to the governing equations, a set of nonlinear ordinary differential equations is obtained in a multiple-degree-of-freedom system. Nonlinear responses of the beam are calculated under periodic lateral acceleration. Nonlinear frequency response curves are computed with the harmonic balance method in a wide range of excitation frequency. Chaotic vibrations are obtained with the numerical integration in a specific frequency region. The chaotic responses are investigated with the Fourier spectra, the Poincare projections, the maximum Lyapunov exponents and the Lyapunov dimension. Applying the procedure of the proper orthogonal decomposition to the chaotic responses, contribution of vibration modes to the chaotic responses is confirmed. The following results have been found: The chaotic responses are generated with the ultra-subharmonic resonant response of the two-third order corresponding to the lowest mode of vibration. The Lyapunov dimension shows that three modes of vibration contribute to the chaotic vibrations predominantly. The results of proper orthogonal decomposition confirm that the three modes contribute to the chaos, which are the first, second, and third modes of vibration. Moreover, the results of the proper orthogonal decomposition are evaluated with velocity which is equivalent to kinetic energy. Higher modes of vibration show larger contribution to the chaotic responses, even though the first mode of vibration has the largest contribution ratio.

Spatially resolved wall temperature measurements during flow boiling in microchannels

Spatial and temporal variations of channel wall temperature during flow boiling microchannel flows using infrared thermography are presented and analyzed. In particular, the top channel wall temperature in a branching microchannel silicon heat sink is measured non-intrusively. Using this technique, time-averaged temperature measurements, with a spatial resolution of 10 μm, are presented over an 18 mm × 18 mm area of the heat sink. Also presented, within a specific sub-region of the heat sink, are intensity maps that are recorded at a rate of 120 frames per second. Time series data at selected point locations in this sub-region are analyzed for their frequency content, and dominant temperature fluctuations are extracted using proper orthogonal decomposition. Results at low-vapor-quality boiling condition indicate that temperatures can be determined from recorded radiation intensities with a temperature uncertainty varying from 0.9 °C at 25 °C to 1.0 °C at 125 °C. The time series data indicate periodic wall temperature fluctuations of approximately 2 °C that are attributed to the passage of vapor slugs. A dominant band of frequencies around 2–4 Hz is suggested by the frequency analysis. Proper orthogonal decomposition results indicate that first six orthogonal modes account for approximately 90% of the variance in temperature. The first mode reconstruction accounts for temporal variations in the dataset in the sub-region analyzed; however the magnitude of fluctuations and spatial variations in temperature are not accurately captured. A reconstruction using the first 25 modes is considered sufficient to capture both the temporal and spatial variations in the data.

Application of autoregressive modeling in proper orthogonal decomposition of building wind pressure

Application of autoregressive (AR) modeling in study of the results of proper orthogonal decomposition (POD) of wind-induced pressure on a low-rise building is described in the paper. The POD principal coordinates (PC) were obtained for point pressure measured simultaneously at 495 taps on the surface of a model of a low-rise building. The AR models were next fitted for the first four PC, and they were compared with the AR models developed for point pressure at representative taps on the building windward wall and roof. It was found that the order of the AR required for the principal coordinates was relatively low, and it did not exceed the order of the AR established for the considered point pressure.