This dataset includes flow conditions, boundary layer parameters, and wall pressure measurements obtained from a turbulent boundary layer experiment conducted over a smooth wall under varying pressure gradients. The measurements were performed using a combination of pointwise microphones and a novel sub-resonant sensor array. The dataset consists of measurements from turbulent boundary layer flows at high Reynolds numbers (6870 < Re_theta < 15,460 and low Mach numbers (0.043 < M < 0.11). The flow conditions include both adverse and favorable mild pressure gradients, characterized by Clause parameter (beta) in the range −0.84 < beta < 0.68. The boundary layer was studied on the wind tunnel sidewall while a NACA0012 airfoil in the test section induced different pressure gradients by varying its angle of attack. For each of the five flow conditions (three Reynolds numbers and three pressure gradient cases, with one common case), the dataset contains: auto-spectral density of fluctuating wall pressure, cross-spectral density function, streamwise wavenumber-frequency spectrum, relevant flow, and boundary layer parameters. The dataset was collected using an in-house developed sensor array, specifically designed to capture sub-convective wall pressure fluctuations with high resolution. Details on the design and development of the array are documented in the referenced PhD dissertation. This dataset is primarily associated with the PhD dissertation: "Measurement and Analysis of Sub-Convective Pressure Fluctuations in Turbulent Boundary Layers: A Novel Methodology." Portions of this research have been published in conference articles, including: "The Characteristics of Sub-Convective Wall Pressure Fluctuations in a Turbulent Boundary Layer" (DOI: 10.2514/6.2024-1904) and "Evaluating Models for Sub-Convective Pressure Fluctuations in Turbulent Boundary Layers" (DOI: 10.2514/6.2024-3048). A note on how this data differs from the data presented in the conference articles is described in the README file associated with the data package. For any inquiries or further details, please refer to the dissertation or contact the corresponding author.
Publisher
University Libraries, Virginia TechLocation
Virginia Tech Stability Wind TunnelCorresponding Author Name
Shishir DamaniFiles/Folders in Dataset and Description
Two types of data formats are available to work with: CSV and MAT. The MAT files package is supported by sample codes that plot the data and also compare it with an existing wall pressure model.
[CSV Files.zip] - Tarball containing data for five flow cases with .csv files
Unzipping should give five folders corresponding to each flow case. The name of the folder corresponds to the following (same as in the dissertation):
• Case I – Reference Re = 0.8E6, AoA = 0 degree
• Case II – Reference Re = 1.2E6, AoA = -10 degree
• Case III – Reference Re = 1.2E6, AoA = 0 degree
• Case IV – Reference Re = 1.2E6, AoA = 12 degree
• Case V – Reference Re = 2E6, AoA = 0 degree
All folders have the same file nomenclature and data organization. The file-based breakdown of data is given below:
• Flow conditions [FlowConditions.csv] - Local parameters of interest in order
o Convective speed – m/s
o Speed of sound – m/s
o Freestream velocity – m/s
o Edge velocity of boundary layer – m/s
o Air density – kg/m^3
o Air viscosity - kg/(m * s)
o Boundary layer thickness – m
o Displacement Thickness – m
o Friction velocity – m/s
o Momentum thickness - m
o Momentum thickness Reynolds number
o Reynolds number per meter
o Displacement thickness Reynolds number
o Freestream Mach number
o Wall shear stress – Pa
• Pointwise spectrum [PointwiseSpectra.csv] – Measured using Brüel & Kjær Type 4138 1/8-inch microphone with a pinhole cap (0.5 mm hole). Column 1 lists the frequency in Hz and column 2 is the spectra density in Pa^2/Hz.
• Averaged double-sided cross-spectral density [CSM.csv] – This was measured using an array of 80 spanwise sensors (5 mm in streamwise direction and 50 mm in spanwise direction) spaced periodically. The data is arranged such that it appends three matrices. Hence, the first set of 544 (rows)x158 (columns) entries represent the frequency grid, followed by streamwise locations in space with the first value corresponding to -0.4290 m and the last value 0.4345 m. This distribution is obtained by conjugate flipping the cross-spectral matrix before Fourier transforming. The spacing between the sensors was 5.5 mm or 0.0055 m. Note that a single column of frequency and a single row of streamwise locations will be unique. The last set of 544x158 entries corresponds to the average double-sided cross-spectral density which is complex-valued. The sensors perform an area average over their rectangular area and this must be kept in mind during comparisons with other data and wall pressure models. This spectral matrix was obtained by considering all possible combinational pairs on the array corresponding to a given streamwise separation. E.g. the density function for the x = 0.055 m, would imply contributions from S1-S11, S2-S12,…S69-S79, S70-S80 assuming streamwise homogeneity.
• Filtered wavenumber-frequency spectrum [Phiqq_filt.csv] – Obtained by Fourier transforming the cross-spectra matrix defined in [CSM.csv] along the streamwise separation dimension. This has a similar format to [CSM.csv] with the first set of 544x80 entries being the frequency grid and the next set being the wavenumbers. A column of the frequency matrix will be unique while a row for the wavenumbers will be unique. The last set of 544x80 entries are the wavenumber-frequency spectra as Pa^2/Hz/(rad/m) complex-valued. Filtering is performed to remove the aliased portion of the spectrum (more details in the dissertation) and the values are set to “0+0i”.
• Calibrated spectra (includes auto and cross-spectrum) [Spectra_cal.csv] – File containing auto-spectral and cross-spectral density as a function of frequency (Hz). The data is arranged so that the corresponding frequency and sensor index are saved. The first three columns read the index as C1 – frequency, C2 – Sensor 1, and C3 – Sensor 2. The first dimension corresponds to frequency and the other two dimensions correspond to the sensor on the array. For e.g., [325+0i 68+0i 12+0i] corresponds with the density function between S68 and S12 for the 325th frequency index. It is to be noted that the density function represents an area-averaged form due to the finite sensor area and not the pointwise density function. To convert the data into a 3D matrix in MATLAB, use:
% Load the CSV
data = readmatrix('Spectra_cal.csv');
% Extract dimensions
row = data(:,1);
col = data(:,2);
slice = data(:,3);
values = data(:,4);
% Determine size
rows = max(row);
cols = max(col);
slices = max(slice);
% Reconstruct the 3D matrix
spectra_cal = nan(rows, cols, slices);
for i = 1:length(values)
spectra_cal(row(i), col(i), slice(i)) = values(i);
End
Hence, spectra_cal(:,i,j) correspond to the cross-spectra between S”I” and S”j”. Also, spectra_cal(:,j,i) is the conjugate flip of spectra_cal(:,i,j). Auto-spectrum is given by spectra_cal(:,i,i) where “i” is the sensor index from 1 to 80 and is real.
==========================================================================
[MAT Files.zip] - Tarball containing data for five flow cases with .mat files and a sample code [Sample_Plotter.m] for plotting.
The filename indicates the Case number (consistent with the reference), the Reynolds number relative to the airfoil chord, and the angle of attack of the airfoil for that flow condition. Eg. CaseIII_ReducedData_1p2MRe_0deg.mat reads flow case III, Re_c = 1.2 million, and AoA = 0 degrees.
The variable breakdown for each file is given below:
• Mean boundary layer parameters - Local parameters of interest
o Air density [AirDensity] – kg/m^3
o Freestream Mach number [Minf]
o Displacement thickness Reynolds number [Redeltas]
o Reynolds number per meter [Reperm]
o Momentum thickness Reynolds number [Retheta]
o Convective speed [Uc] – m/s
o Edge velocity of boundary layer [Ue] – m/s
o Freestream velocity [Uinfty] – m/s
o Speed of sound [c0] – m/s
o Boundary layer thickness [delta] – m
o Displacement Thickness [deltas] – m
o Momentum thickness [theta] - m
o Wall shear stress [tauw] – Pa
o Friction velocity [utau] – m/s
o Air viscosity [viscosity] - kg/(m * s)
• Pointwise spectrum [Gpp_point] – Pa2/Hz {real} – In matrix format with rows match frequency vector stored in [freq_red] and columns matching number of sensors (80).
• Averaged double-sided cross-spectral density [cross_full_avg] – Pa^2/Hz {complex} – The description is given previously and it size is 544x158 with first index matching frequency vectore [freq_red] and second index matching the [streamlocs] but with negative separations as well defined by [-flip(streamlocs(2:end-1)) streamlocs].
• Filtered wavenumber-frequency spectrum [Phiqq_filt] – Pa^2/Hz/(rad/m) {complex} – Obtained by Fourier transforming the [cross_full_avg] variable along the streamwise separation dimension. A filtering is performed to remove the aliased portion of the spectrum (more details in the dissertation) and the values are set to “0+0i”. Associated wavenumbers are stored in [kstream] and frequency in [freq_red].
• Calibrated spectra (includes auto and cross-spectrum) [spectra_cal] – Pa^2/Hz (complex) - File containing auto-spectral and cross-spectral density as a function of frequency (Hz) in a 3D array. The first dimension corresponds to frequency and the other two dimensions correspond to the sensor on the array. Hence, spectra_cal(:,i,j) correspond to the cross-spectra between S’i” and S”j”. Also, spectra_cal(:,j,i) is the conjugate flip of spectra_cal(:,i,j). Auto-spectrum is given by spectra_cal(:,i,i) where “i” is the sensor index from 1 to 80 and is real. It is to be noted that the density function represents an area-averaged form (implies filtering of high frequencies) due to the finite sensor area and not the pointwise density function.
==========================================================================
Note on differences between data presented in DOI references and this dataset:
The data presented in the conference articles were corrected for two reasons: (i) aliasing observed due to an effective spanwise shift from the weighted sensitivity of the sensor, and (ii) a phase correction applied to a non-linear behavior observed in the phase between adjacent sensors. Further investigation of the corrections revealed high uncertainty in the sub-convective domain of the wavenumber-frequency spectrum. This is observed in Figure 13 (10.2514/6.2024-1904) throughout the sub-convective domain and Figure 7 (10.2514/6.2024-3048) at frequencies below 1000 Hz. Hence, the current dataset does not include these corrections revealing more accurate flow behavior and the aliased portions in the wavenumber-frequency spectrum have been omitted by performing a thorough analysis of the weighted nature of the sensor and modeling its effects on the streamwise wavenumber-frequency spectrum (further described in the PhD dissertation, section 7.4.2).
=============================================================================
Details on how to compare the data with existing wall pressure models:
The measured data represents the spectral density functions as area-averaged by finite area sensors. The sensors are 5 mm in the streamwise direction and 50 mm in the spanwise direction with a rectangular surface area. The area-averaging character is assumed to be uniform over this area and uses the convolution between the true pressure spectrum and the area sensitivity of the sensor the measured pressure spectrum for a finite area sensor as described by various studies like Kirby (1969) and Blake (2017). A sample code [Sample_WPModelCompare.m] is provided as part of the [MAT Files.zip] package describing how to compare the measured cross-spectral density function and the wavenumber-frequency spectrum with the Modified Corcos model. This code can be adapted to other wall pressure model definitions.
