This paper investigates the effect of training data on the accuracy of turbulent flow predictions in the wake region of a linear turbine cascade using physics-informed neural networks (PINNs). While it is well known that PINNs can solve the unclosed Reynolds-averaged Navier-Stokes (RANS) equations when sufficient training data are available, the specific characteristics of the data such as the quantity and location required for accurate predictions remain largely uncertain. To explore this, a PINN is constructed to solve the unclosed compressible RANS equations leveraging training data from a computational fluid dynamics (CFD) solution of a turbine blade. The training data are then selectively sampled with future optical test campaigns in mind. This sampling includes varying the pitchwise surveys with evenly spaced training points, randomly sampled points, and CFD-guided sampling. For each case, the PINN is trained on data for the velocity components, temperature, pressure, and Reynolds stresses. Good agreement is seen between the PINNs-predicted quantities and the CFD solution, even in areas where the PINN was not provided training data and excellent agreement in regions where data were provided. It is shown that the PINN can provide acceptable solutions to the unclosed compressible RANS equations with roughly 100 data points downstream of the blade, and even better results when provided with roughly 200 points. This shows promise for future test campaigns which seek to combine AI-based tools with experimental techniques.

A versatile procedure is presented for simulating the growth and thermal impact of deposition in representative gas turbine cooling circuits. The methodology comprises the following sequence: (1) employ computational fluid dynamics (CFD) to obtain a steady flow solution, (2) predict particle impacts and deposits with the Ohio State University (OSU) deposition model, and (3) implement conjugate mesh morphing of both the fluid and solid domains to simulate the evolution of deposit structures. Impinging jet experiments are conducted at 990 K with aerosolized Arizona Road Dust (ARD 0-10 & micro;m) impacting an unheated target. The simulations accurately model deposit cone evolution. A "regridding" step is further employed in the mesh-morphing routine to significantly improve stability during adaptation of the mesh. Subsequent tests are conducted in an effusion cooling deposition facility, where an effusion plate is heated to 1144 K with cooling air supplied at varying temperatures (811-978 K) at a constant backflow margin of 3%. ARD (0-10 & micro;m) is injected, and the reduction in cooling air mass flow is monitored. The results illustrate that higher cooling air temperatures lead to a quicker reduction in cooling air mass flow and variations in the deposit structure within the effusion holes. The enhanced mesh-morphing routine is integrated with the temperature-sensitive OSU deposition model to simulate these tests. The simulations capture the increasing blockage rate with temperature and display similarities in the deposit structures. The "regridding" step is determined to play a critical role in ensuring stability of the mesh-morphing routine in effusion geometry.

A new deposition cooling facility was designed and constructed to examine the time evolution of dust deposition and its effect on the cooling effectiveness of internal cooling circuits typical of gas turbine hot sections. The facility utilizes representative gas turbine engine cooling geometries and operates at relevant coolant and hot gas flow temperatures (650 K and 1300 K, respectively). A combustion chamber is used to create a high-speed (180 m/s) 2 cm diameter jet that impinges at a 26 deg angle (from the horizontal) onto a pre-oxidized flat plate made from a nickel-based alloy. The backside of this plate is cooled by an array of 0.5 mm diameter impingement cooling jets with a 1.03 pressure ratio relative to the hot jet discharge pressure. An infrared (IR) camera is used to record the surface temperature of the flat plate as airborne dust is delivered through the cooling circuit. The dust delivered to the test article is 0-5micron Arizona Road Dust. Deposit evolution and cooling effectiveness trends were observed by performing successive tests with varying dust delivery periods while maintaining constant dust loading (concentration). Observable patterns in the deposit behavior corroborate quantitative measurements of changes in local cooling effectiveness as measured by the IR camera. A limited test series at lower coolant and jet temperatures highlights the role of temperature in the deposition process.

The effects of electrostatics on the deposition of 0-5 & micro;m Arizona Road Dust within a gas turbine hot section were studied. These experiments focused on different cooling methods found in turbine components: impingement cooling jets and double-walled cooling circuits. Particles within a flow can experience tribocharging through collisions with their surroundings: walls of passages and other particles. For the impingement cooling jet, dust was entrained with hot air through a 6.35-mm-diameter pipe at a mean velocity of 57 m/s. The gas temperature ranged from 755 K to 1005 K to test the temperature sensitivity of any electrostatic effects. The dust-laden jet impinges on a charged plate (-12 kV to +12 kV) to form a deposit. The effect of plate charge on the morphological characteristics of the deposit formation was assessed. For the double-walled cooling circuit, a charge (-5 kV to +5 kV) was applied to both the effusion interior and impingement exterior. Dust was entrained in an 810 K flow with a pressure ratio (PR) of 1.03. No significant electrostatic effect was observed for this geometry. These findings were supported by lower-temperature gravity-drop tests through charged parallel plates and a charged cylindrical pipe. The more fundamental tests confirmed that Arizona Road Dust particles tribocharge negatively and are attracted to positive potential surfaces. This effect appears to be augmented at higher temperature. The cylindrical gravity-drop confirmed that electrostatic influence diminishes due to the absence of electric field lines within enclosed geometries.

A novel active flow control (AFC) technique is applied to a 30 degrees swept, NACA 643-618 wing in a Re = 100,000 flow. Three wings of the same geometry were equipped with three different flow control configurations. Two of the wings (wing 60 & 80) were equipped with a single row of twenty VGJs skewed 60 degrees from the freestream air and pitched 30 degrees from the local wing surface. The third wing (multi-fence) was equipped with four rows of five VGJs at 20, 40, 60, and 80% span oriented in the same manner. Global force data showed the multi-fence configuration increased lift up to 15.66%, 15.08%, and 12.84% at momentum coefficients of 2%, 1.5%, and 1%, respectively. The center of pressure was shifted forward, creating more positive pitching moment coefficients than the wings equipped with a single spanwise location of AFC. Drag was also reduced in the near stall regime. Surface oil flow visualizations (SOFV) revealed that actuation at 20% on the multi-fence wing eliminated the presence of an LEV near the wing root for light stall and reattached flow. Compared to actuating at a single spanwise location, multi-fence AFC acted as multiple decoupled flow enhancements. At low, increasingly practical C-mu, location was more important than mass flow rate in determining impact. As such, distributed AFC was found to be more effective than single location actuation.

Previous studies have demonstrated the effectiveness of using fluidic actuators to throttle the flow through a nozzle guide vane, enabling a variable area turbine (VAT) with no moving parts. A simplified model was developed to assess the driving design factors that influence the blocked flow fraction and throttling effectiveness of the actuators. The model approximates slot injection using a simplified stream tube analysis, neglecting mixing with the primary passage flow. Throttling performance is predicted for fluidic actuators with different widths, orientations, streamwise locations, and pressure ratios. This simplified model evaluates these actuators in a constant width straight duct and a variable width duct. Two-dimensional CFD of injection in the nozzle guide vane passage reported strong agreement with the results of the simple model using the variable width duct. Higher injection pressure ratios result in more injected mass flow but not significantly higher throttling effectiveness. Wider slots retain the effectiveness and block much more primary flow. Upstream oriented injection proves best for both effectiveness and blockage capabilities. Finally, there is an optimal injection location relative to the passage throat, which is extracted by both the model and CFD. This model has potential application for optimizing fluidic VAT throttling designs while saving on the cost of high-fidelity testing and production of these configurations.

The effectiveness of partial-span, dual-slot trailing-edge (TE) Coanda actuators on swept wings remains uncertain in the hover regime of rotor-blown wing tailsitter UAVs. This study investigates such effects using a benchtop setup with a vertically mounted wing in propeller-driven flow. Force measurements were collected via an internal load cell across sweep angles ranging from 0 degrees to 30 degrees. For the unswept wing, dual-sided blowing-relative to single-sided blowing-improves linearity in the lift response, increases thrust vectoring effectiveness, and delays stall, albeit with increased mass flow demand depending on the chosen blowing strategy. Background-oriented Schlieren visualization reveals a correlation between stall onset and the Coanda jet turning angle exceeding 90 degrees. As wing sweep increases, lift and pitching moment responses become increasingly asymmetric between dual-sided upward and downward blowing directions. This is attributed to lateral deviation of the propeller slipstream as the wing is swept, as confirmed by surface oil flow visualization. These results highlight complex propeller-wing-actuator interactions in swept-wing tailsitter during hover and emphasize the need for tailored blowing strategies in future UAV designs employing TE Coanda actuation.

An impingement/effusion cooling configuration was studied in a high pressure deposition facility with two different dust types, Arizona Road Dust (ARD) and Air Force Research Lab (AFRL05). Each dust had a size distribution of less than 10 microns. The results of the current study compare the effects of cycling on deposition as evidenced by capture efficiency and discharge coefficient. Conducting the current study involved the dust entering the impingement/effusion cooling geometry from a plenumized condition (e.g. through an upstream dump diffuser). Comparisons are made to an alternate test fixture where dust-laden air is presented to the test plates along a path parallel to the surface. All tests were conducted in a high-pressure deposition facility that simulates the environment of a combustor liner or turbine vane cooling circuit. Coolant flow temperature conditions reached around 810 K (1000 degrees F) and discharged into a 15 atm (230 psi) cavity pressure with a nominal pressure ratio of 1.027. Once conditions were reached dust was delivered in 1 to 2 cycles while keeping the total amount of desired dust delivered constant (roughly 8 grams). When testing more than 1 cycle, a dwell period of 24 hours at room temperature occurred prior to the second delivery period. In the dump diffuser configuration both dusts resulted in less deposition-induced blockage compared to the parallel flow as seen in the discharge coefficient behavior. Additionally, the ARD dust proved to be much more aggressive than the AFRL05 dust in both diffuser configurations. The dump diffuser setup did not show any cycling effect on deposition with AFRL05 but cycling does have an effect on deposition for the ARD dust.

Experiments were conducted in a high-pressure deposition facility to study the effect of engine cycling on the successive buildup of deposits as aircraft shut down and power up. The test facility simulates the pressure and temperature environment of a combustor liner or turbine vane cooling circuit. The coolant flow temperature reaches 894 K (1150 degrees F) and discharges into a 17-atm (250-psi) cavity pressure at a nominal pressure ratio of 1.027. AFRL05 test dust with a 0- to 10-mu m size distribution is added to the coolant gas stream. The cooling circuit consists of a double-walled impingement/effusion cooling plate with nominal hole sizes on the order of 0.5 mm. To simulate cycling, the facility is brought up to the desired operating conditions where the first batch of dust is delivered (2-8 g). The facility is then ramped down to ambient conditions. Following a "dwell" period of approximately 24 h, another batch of dust is delivered once the facility is brought back up to the desired operating condition. Test data were acquired for one, two, and four cycles with different dust masses delivered. Values such as discharge coeffi- cient, pressure ratio, Reynolds number, and temperature are used to evaluate the effects of dust deposition on the impingement/effusion plate setup. Dust capture efficiency is shown to be insensitive to cycling whereas flow blockage is negatively impacted by an increased number of cycles for the same total mass delivered. The sloughing of deposit structures during cooling circuit cool down is postulated to be responsible for the observed behavior.

Previous experiments demonstrated the effectiveness of using fluidic actuators to throttle the core mass flow through a nozzle guide vane passage. This enables a variable area turbine (VAT) without mechanically moving vanes. The most promising of the fluidic actuators tested was a slot on the suction side of the nozzle guide vane. Using a simple model as a guide, a two dimensional, RANS, CFD Design of Experiments (DOE) was completed to evaluate the predicted mass flow rate reduction, effectiveness, pressure loss, and exit flow angle for various slot geometries. This DOE was succeeded by the experimental evaluation of seven candidate slot geometries in a small-scale, single nozzle guide vane passage. These candidate slots span the geometric parameters previously described. Experiments showed that when the injection pressure matches the upstream pressure, increasing slot width from 6% to 19% of the geometric throat caused the percentage of primary flow blocked to go from 6.4% to 24.8%. When the slot was pitched from 0 to 60 into the oncoming flow the percentage of primary flow blocked went from 16.0% to 18.2%. Finally, when the slot location was moved away from the computationally determined optimal location the primary flow blocked was reduced. All slot geometries tested had promising effectiveness, generally blocking around twice as much mass flow as they were injecting. Additionally, total pressure loss was seen to strongly correlate with the amount of flow blocked no matter the slot geometry. The trends seen in the simple model and CFD did match the experiments, however CFD appears to have overestimated the performance of the best slot designs.

Previous experiments demonstrated the effectiveness of using fluidic actuators to decrease the core mass flow through a nozzle guide vane. An investigation was conducted into increasing the mass flow through a nozzle guide vane using fluidic actuators. Coupling actuators that increase mass flow with actuators that decrease mass flow allows for a Variable Area Turbine (VAT) that has dual function throttling control without mechanically moving vanes. The dual function geometry investigated includes a cavity on the suction side of the vane placed at the location of the geometric throat, and injection mass flow from a Coanda actuator placed at the upstream side of the cavity blowing downstream and into the cavity. With no injection, the primary core flow passes over the cavity, resulting in an effective throat width that is similar to the original geometric throat width. When the fluidic actuation is turned on, the injection flow entrains the core flow into the cavity, expanding the flow, and increasing the effective throat width. Using two-dimensional, periodic, RANS CFD, a design of experiments was conducted to evaluate the effectiveness of the Coanda actuator to increase mass flow for varying cavity and injection geometries, while also evaluating the effectiveness of a blocking actuator to reduce mass flow. The optimum case showed the ability to increase the primary mass flow by 15% and decrease primary mass flow by 19.9%, with the effectiveness of the actuators being approximately two in both cases. This optimum geometry case was then experimentally tested in a small-scale single nozzle guide vane passage. The experimental results show that while the blocking actuator was able to reduce mass flow effectively, the Coanda actuator was not able to increase the mass flow as effectively as predicted. Explanations for this disparity between the CFD results and experimental results are discussed.

The effectiveness of a micro-cavity near the leading-edge of sinusoidally pitching NACA 0012 airfoil is experimentally investigated as a means of dynamic stall control. The microcavity is sized and positioned to excite resonant oscillations during deep stall and thus collapse the leading-edge separation bubble. Wind tunnel testing focused on freestream conditions of M = 0.20 - 0.25, Re = 6.8 - 8.35 x 10(5) and reduced frequency, k = 0.039 - 0.049. Micro-cavity results at M = 0.20 and 0.25 showed improved lift, reduced hysteresis, and earlier downstroke flow reattachment relative to the uncontrolled airfoil. Normal and axial force contours revealed the presence of intermittent pitching cycles of beneficial control, which were increasingly present in the downstroke portions of M = 0.25 flow. These events, which are indicative of effective micro-cavity control, were separated and conditionally averaged to reveal significant (up to 56%) reductions in hysteresis and a transition from negative to positive aerodynamic damping. This study offers the first experimental verification of the micro-cavity actuator for dynamic stall control.

Experiments were conducted to determine the effects of electrostatics on the deposition of airborne dust particles within a gas turbine hot section. One experiment simulates particle laden flow through an impingement cooling jet. Dust (0-5 mu m Arizona Road Dust) was entrained with hot air through a 3.66 m long, 6.35 mm diameter tube at a mean velocity of 57 m/s. The gas temperature ranged from 755 K to 1005 K to test temperature sensitivity of any electrostatics effect. As the dust travels through the pipe, it collides with the pipe's walls and other dust particles resulting in fribocharging. The dust-laden jet impinges on a charged plate (-12 kV to +12 kV) where a deposit is formed. The effect of plate charge on the deposit formation was assessed pertaining to its size, shape, and volume. A second experiment was performed to determine the effect of electrostatics on deposition in a double-walled cooling circuit. A charge (-5 kV to +5 kV) was applied to both the effusion interior and impingement exterior enclosures. Dust (0-5 mu m Arizona Road Dust) was entrained in an 810 K flow with a pressure ratio (PR) of 1.03 to simulate gas turbine cooling flow operating conditions. No significant electrostatic effect was observed for this geometry. These two results at relevant temperatures for gas turbine cooling were corroborated by additional gravity drop experiments at lower temperature: through charged parallel plates and through a charged cylindrical pipe. These more fundamental tests confirmed the finding that ARD particles tribocharge negatively and are thus attracted to positive potential surfaces. This effect appears to be augmented at higher temperature. Also, the cylindrical pipe gravity drop results confirmed that when particles are in an enclosed geometry, no electrostatic effect is observed due to the lack of electric field lines.

Active flow control (AFC) is applied to a 45 degrees swept wing in low-speed flight conditions via a fluidic fence created by a streamwise row of vortex generating jets (VGJs). Computational fluid dynamics (CFD) simulations are utilized to evaluate many geometric parameters of control when the wing is at 15 degrees and 20 degrees angle of attack and Re = 75,000. While the pitch angle (phi) of the VGJs is held constant at 30 degrees, the skew angle of the VGJs is evaluated at three orientations: inboard (theta = -60 degrees), outboard (theta = +45 degrees), and zero-skew (theta = 0 degrees). The fence is tested at three locations: 30%, 50%, and 70% semispan. The nominal VGJ fluidic fence contains 20 VGJs, and in this study, the size (streamwise extent) of the fence is evaluated in different groupings of jets: 5, 10, and 20 jets. Control is applied with 2% momentum coefficient (C-mu) when the full row is utilized, but with partial rows, the mean jet velocity is held constant, reducing the momentum coefficients to 1% and 0.5% for 10 and 5 total jets, respectively. By varying the geometric parameters of the fluidic fence in simulations, configurations that maximize the ratio between lift coefficient increment and required momentum coefficient are determined. The optimal configurations from simulation results are evaluated experimentally across a wide range of angles of attack (alpha = 0 degrees-38 degrees) with the same freestream conditions as in the CFD and compared on the basis of power coefficient (C-pi).

A novel fluidic oscillator design, in which the oscillator is curved along the primary flow direction, was evaluated. The effects of mass flow rate, hydraulic diameter, aspect ratio, curvature radius, surface roughness, inlet orientation, and nonconstant curvatures were experimentally investigated. Measurements of the oscillation Strouhal number, discharge coefficient, spreading angle, and sweeping angle were used to characterize the jet. The resulting oscillators were found to fall into one of three categories: oscillated at the same frequency and sweep angle as conventional flat oscillators; oscillated at a slightly higher frequency and lower sweep angle than flat oscillators; or no dominant oscillation frequency detected and with no sweeping action. An unsteady Reynolds-averaged Navier–Stokes computational fluid dynamics simulation revealed fundamental differences in the internal flow mechanisms between flat and curved oscillators that drive the sweeping jet. The curvature between 37.5 and 62.5% of the total length, or the region from the inlet nozzle to halfway through the main chamber, was a primary factor influencing the response type of a design. Due to the curvature of these oscillators, they have the ability to be used in geometrically constrained spaces, such as the leading edge of wings and turbine vanes.