Dual-Tomographic Hybrid-PIV/PTV for extracting quantitative spatio-temporal structure of a turbulent jet

Many flow studies of both practical and fundamental interest permit dimensional reduction through simplifying assumptions such as spatial symmetry or temporal averaging, enabling the use of accessible techniques like planar 2C2D PIV and hotwire anemometry. However, accurate description of flows with complex three-dimensional geometry or significant temporal fluctuations demands more comprehensive kinematic measurements. This need has driven development of stereo-PIV (Buchner et al. (2012)), tomographic PIV (Soria (1996)), and holography (Sun et al. (2020)), which capture sufficient additional dimensions to infer the underlying multidimensional dynamics of such flows. However, acquiring data with temporal information encoded necessitates large reductions in spatial extent and/or resolution, greatly restricting the extraction of a flow’s full 3-D spatio-temporal dynamics. Here, a dual-tomographic PIV system is presented which bypasses these sacrifices by running two independent systems concurrently with a discrete time offset, encoding temporal dynamics at a resolution satisfying the Nyquist criterion up to O(10 kHz) at high spatial resolution over a substantial physical domain. (Sikroria et al. (2022)) To validate the method, a simulation designed to match the characteristics of the real M ≈0.9 jet experiment is carried out. Resolution and scaling is matched to the real PCO Panda 26 MPx DS cameras, which acquire 2-D image samples with a pixel size of approximately 10 µm. This allows the reconstruction of full 3-D flow fields with ∼100 billion voxels, spanning approximately 10 diameters downstream of the 5 mm jet exit. The dual-system configuration additionally de-correlates ghost particles between frames, mitigating the primary source of bias error in tomographic reconstructions. A computationally efficient hybrid PIV/PTV tracking scheme is proposed that initialises particle trajectories from the first two frames via full 3-D reconstruction and 3-D cross-correlation, then continues tracking in the image plane for frames 3 and 4, avoiding the most expensive processing stages entirely. To quantify algorithm performance, synthetic particle fields are generated by advecting particles through velocity fields from a high-resolution DNS of a Mach 0.6 turbulent round jet at Re = 10,000, with virtual cameras placed at experimentally equivalent positions. A parametric sweep over seeding density and search radius demonstrates robust track recovery up to 100k particles, with correct-to-incorrect track ratios remaining strongly favourable at this seeding level, establishing practical operating bounds for the experimental configuration.