Limits and accuracy of the Stereo-LFC PIV technique and its application to flows of industrial interest

Abstract

Particle image velocimetry with local field correction (LFC PIV) has been tested in the past to obtain two components of velocity in a two dimensional domain (2D2C). When compared to conventional correlation based algorithms, this advanced technique has shown improvements in three important aspects: robustness, resolution and ability to cope with large displacements gradients. A further step in the development of PIV algorithms consists in the combination of LFC with the stereo technique, which is able to obtain three components of velocity in a plane (2D3C PIV). In this work this combination is implemented and its performance is evaluated carrying out the following two different tasks:

• – Comparison of robustness and accuracy for large and small scale flow structures. This is carried out using three techniques: the conventional Stereo PIV, the Stereo-LFC PIV and the Stereo-Multigrid PIV enhanced with image distortion.

• – Insight on the limit of resolvable scales for the Stereo-LFC. This task is relevant because the resolution attainable by this combination is higher than what has been obtained by the rest of the herein used algorithms.

The first task has been achieved using synthetic images. Afterwards the coherence of the results has been checked with real images. The results show improvement of Stereo-LFC PIV in respect to Stereo-Multigrid PIV enhanced with image distortion. The performance of Stereo-LFC when only large scales are involved shows an increase of the dynamic range of measurable vorticity. When small scales are analysed, the magnitude of the error resulting when using Stereo-LFC is about half of the one obtained for the Stereo-Multigrid measurements. Results with errors below 20% have been achieved for some of the cases with peak vorticities as large as 1.8 Δt ⁻¹ (in the absence of out-of-plane displacements), out-of-plane loss of particle pairs of 65% (with a low peak vorticity of 0.06 Δt ⁻¹) and peak vorticities as large as 1.5 Δt ⁻¹ with 50% particle pair loss. For the second task most of the information has been obtained using real images. It has been found that the resolution limit is very dependent on the robustness of the algorithms against image defects and variability. The results show a remarkable improvement when using the Stereo-LFC PIV processing, although a full quantification and characterization would need further study because of the variety of noise sources possible in a real image.