Adaptive image interrogation for PIV

Abstract

As an experimental tool, Particle Image Velocimetry has quickly superseded traditional point-wise measurements. The inherent image processing has become standardized though the performances are strongly dependent on user experience. Moreover, the arduously selected image interrogation parameters are applied uniformly throughout the image snapshots and image sequence but seldom comply with the observed fluid’s convective motion, spatial distribution in length scales or signal distribution. Instead, a degree of adaptation in the image analyses is required to estimate the velocity field underlying the image recordings as accurate as possible and preferably within an automated fashion. In this work, the aim has been a global solution which through adaptivity of the interrogation parameters (window size, eccentricity, orientation, location and overlap) remains adequate in the majority of encountered problems. This dissertation proposes to go in line of a recursive approach autonomously adapting to both signal and flow conditions. Correlation window location, number and size are regulated taking into account seeding quantity and flow fluctuation magnitude. Signal quantization is based on individual particle image segmentation while spatial variance in velocity served as a heuristic for flow adaptation. The new interrogation method surpasses the compromise between spatial resolution and robustness and places more and smaller windows where the flow requires it and seeding allows it. Vice versa, less of these unnecessary small windows are placed in regions where the flow does not require it (i.e. absence of gradients or fluctuations in velocity). A variant of the spatially adaptive interrogation method is proposed that refines window size, shape, orientation and spatial distribution based on the ensemble averaged velocity field and image properties. The use of ensemble averaged properties enables the reliable application of non-isotropic resolution in contrast to the instantaneous adaptive approach where the latter is impracticable. This approach additionally allows to reduce the number of interrogation windows without overly compromising the measurement spatial resolution where needed. To cope with typical problems of PIV near interfaces, an innovative interface treatment has been proposed incorporating wall adaptivity in an automated manner by gradually increasing the sampling rate in the vicinity of the wall, rotating the correlation windows parallel to the interface and reducing wall-normal window sizes. The enhanced performances of the adaptive interrogation approach have been extensively assessed and demonstrated on a large basis of experimental flow image recordings.