3D Interrogation Algorithms

Abstract

The interrogation of the volumetric distributions of light intensity obtained in output from the tomographic reconstruction is a crucial aspect, both for the computational cost and the accuracy. The 3D interrogation algorithms have to be optimized to handle large amount of data and, simultaneously, to exploit the more complete information on the flow structure to increase the accuracy. The computational cost reduction strategies are discussed in Sec. 2. Three complementary approaches are presented. The first approach is based on optimization of the 3D cross-correlation interrogation using a software-binning based algorithm in the predictor estimation (UNINA, DLR(C) and LAVIS), optimization of the data treatment to reduce redundant calculation in case of overlapping windows (UNINA), and sparse-aided algorithms (UNINA, UNIMO). The second approach consists in the data reduction from 3D to 2D summing neighbouring planes of the reconstructed volumes (DLR(C) and IOT). The third strategy is based on advanced parallelization on GPUs (IOT, LAVIS), or using OpenMP-based software (IOT, DLR(C), UNINA, PPRIME). The Integration of the interrogation algorithms in the “DaVis” commercial software is performed by LAVIS. The strategies to increase the accuracy and reliability of the interrogation algorithms are discussed in Sec. 3. A theoretical study of the modulation transfer function of the cross-correlation algorithm and of the negative effects due to poor discretization of weighting windows in the cross-correlation step (UNINA) is reported. The possibility to calculate the full velocity gradient opens the field to new adaptive PIV strategies based on properly shaped weighting windows (TUD). A great margin of improvement of 3D PIV is provided by time-resolved imaging. This aspect is discussed in Sec. 4. In particular, a solution to increase the measurement accuracy using properly composed correlation maps from multiple exposures is presented (TUD). Furthermore, the tracking of the particles over time can be used to measure the Lagrangian acceleration (TUD). Finally, an alternative approach based on optical flow methods (PPRIME) is provided in Sec. 5.