How it works
Raw ADV records turns each Vectrino record into one row of statistics: it reads the file, rejects bad samples, removes spikes, and computes the mean velocities, turbulence intensities, turbulent kinetic energy and Reynolds shear stresses. Survey maps takes a table of such rows with their positions and draws each quantity across the channel, with the headline numbers of the wake.
Files it reads
- Vectrino and Vectrino Plus ASCII exports (
.dat), one row per sample. The column order is taken from the.hdrfile written alongside; without it, the layout is recognised from the column count (status, four velocities, four amplitudes, four SNR and four correlation columns, after an optional time and sample counter). - CSV time series with columns such as
u,v,win m/s, cm/s or mm/s, and optionallytime,snr,cor. - Node tables: one row per point with
x(m),y(m)and any ofU_mean (m/s),TKE (m2/s2),<u'v'> (m2/s2)and so on. Rows whose point id ends in_mirrorcount as mirrored, not measured.
Quality control
- Samples with a signal-to-noise ratio or beam correlation under the thresholds are rejected, judged on the weakest beam or on the beam average. Standard settings: SNR of at least 15 dB and correlation of at least 70 %.
- Spikes are found by phase-space thresholding (Goring and Nikora 2002) with median and MAD spreads (Wahl 2003): each sample, its first difference and its second difference must stay inside an ellipsoid sized by the universal threshold √(2 ln N). The test repeats until it finds no new spikes. Clipping at k standard deviations is the simpler alternative.
- Statistics use the kept samples only. Rejected samples are bridged by linear interpolation so that spectra see an evenly sampled record.
- The Seminar 2026 settings (no SNR or correlation cut, 4σ clipping, W as the mean of Z1 and Z2, 50 Hz) reproduce the node table behind the 2026 seminar: median differences over the 205 raw records are 0.01 % for U and 0.5 % for k.
Statistics
Spectra use Welch's method: Hann window, half-overlapping segments of 2n samples close to N/8, linear detrending. The dashed line has the −5/3 slope of the inertial range.
Maps
- Each quantity is interpolated linearly on a Delaunay triangulation of the points. Fill to the side walls extends every station to both walls with zero gradient and smooths lightly (12 mm); the areas it adds are extended from the nearest measurements, not measured. With the fill off, cells farther than the chosen distance from any point stay blank.
- Mirroring maps y to B − y and reverses the sign of V, ⟨u′v′⟩ and ⟨v′w′⟩. Mirror stations measured on one side adds only the missing half of those stations, and never replaces a measured point. Symmetric replaces the whole upper half by the mirror image of the lower half, for symmetric presentation maps.
- U₀ and k₀ are means over the approach window (the first 0.2 m of the survey). Velocity deficit = 1 − Umin/U₀ behind the piers, Froude number Fr = U₀/√(gh), pier Reynolds number ReD = U₀D/ν with ν = 10⁻⁶ m²/s, blockage = nD/B.
The same steps in Python
The advwake Python package runs the same quality control and statistics from the command line and draws publication figures as PNG, SVG and PDF.
advwake process raw/ --stations stations.csv -o results advwake figures results/node_results.csv -o figures
About the example data
The examples on this page are synthetic: a simple wake model of five cylinders (D = 0.10 m, S/D = 2) across a 1 m wide flume, and Vectrino-style records generated with known statistics, spikes and short low-correlation bursts so that the quality control has work to do. They show what the tool does; they are not measurements. Replace them with your own files at any time.
ADVwake is developed by Soumyajit Khatua, B.Tech Civil Engineering, Odisha University of Technology and Research (OUTR), Bhubaneswar.
References
- Goring, D. G., and Nikora, V. I. (2002). Despiking acoustic Doppler velocimeter data. Journal of Hydraulic Engineering, 128(1), 117 to 126.
- Wahl, T. L. (2003). Discussion of "Despiking acoustic Doppler velocimeter data". Journal of Hydraulic Engineering, 129(6), 484 to 487.
- Welch, P. D. (1967). The use of fast Fourier transform for the estimation of power spectra. IEEE Transactions on Audio and Electroacoustics, 15(2), 70 to 73.