No abstract

No abstract

Uncertainty quantification in planar particle image velocimetry (PIV) measurement is critical for proper assessment of the quality and significance of reported results. New uncertainty estimation methods have been recently introduced generating interest about their applicability and utility. The present study compares and contrasts current methods, across two separate experiments and three software packages in order to provide a diversified assessment of the methods. We evaluated the performance of four uncertainty estimation methods, primary peak ratio (PPR), mutual information (MI), image matching (IM) and correlation statistics (CS). The PPR method was implemented and tested in two processing codes, using in-house open source PIV processing software (PRANA, Purdue University) and Insight4G (TSI, Inc.). The MI method was evaluated in PRANA, as was the IM method. The CS method was evaluated using DaVis (LaVision, GmbH). Utilizing two PIV systems for high and lowresolution measurements and a laser doppler velocimetry (LDV) system, data were acquired in a total of three cases: a jet flow and a cylinder in cross flow at two Reynolds numbers. LDV measurements were used to establish a point validation against which the high-resolution PIV measurements were validated. Subsequently, the high-resolution PIV measurements were used as a reference against which the low-resolution PIV data were assessed for error and uncertainty. We compared error and uncertainty distributions, spatially varying RMS error and RMS uncertainty, and standard uncertainty coverages. We observed that qualitatively, each method responded to spatially varying error (i.e. higher error regions resulted in higher uncertainty predictions in that region). However, the PPR and MI methods demonstrated reduced uncertainty dynamic range response. In contrast, the IM and CS methods showed better response, but under-predicted the uncertainty ranges. The standard coverages (68% confidence interval) ranged from approximately 65%-77% for PPR and MI methods, 40%-50% for IM and near 50% for CS. These observations illustrate some of the strengths and weaknesses of the methods considered herein and identify future directions for development and improvement.

No abstract

A moving wall section attached to a piezoelectric actuator was used to perturb the airflow exiting a fully developed turbulent channel. The Reynolds number based on the channel width and centerline velocity was 4240. The maximum velocity of the moving wall section was 9.5 cm/s ͑2.3% of the mean centerline velocity͒, and the maximum displacement was 120 microns, corresponding to 1.84 wall units. The actuator frequency and displacement amplitude were tuned independently to generate different effects on the flow, and both quantities were documented to provide precise boundary conditions for numerical codes. Hot-wire measurements showed that actuation affects both the mean and rms velocity profiles downstream of the channel exit. In all cases, forcing yields mean profiles that are symmetric with respect to the centerline. However, forcing at low frequencies (Stр0.30) causes faster decay of the centerline velocity, higher spreading rates in the far field, and asymmetric rms profiles compared with unforced flow. The maximum rms value crosses from the actuator side to the nonactuator side as the flow moves downstream. This behavior is thought to be caused by staggered but asymmetric vortical structures developing in the opposing shear layers. Forcing from Stϭ0.39 through 1.46 leads to altered but symmetric rms profiles and spectra compared with unforced flow. Forcing in the range StϽ0.50 yields centerline rms values that initially are larger than, but further downstream smaller than in the unforced flow.