Abstract: In recent works, the authors have reported an extension of PIV, called "Interface Gradiometry" (PIV/IG) that, by shearing templates to various degrees, directly measures the instantaneous velocity gradient at a fixed wall. This approach works well when the height of templates is carefully chosen. Our recent generalization of the original PIV/IG (denoted PIV/IG+; Nguyen & Wells 2006), does not require a template height to be specified a priori for pattern matching, and is thus more widely applicable. It also introduces an image transformation to handle curved boundaries. The present communication considers the sensitivity of PIV/IG+ results to errors in estimated wall position.
Introduction
Measuring fluid behavior in the immediate vicinity of a solid surface—the "near-wall region"—presents significant challenges for experimental fluid dynamics. Standard Particle Image Velocimetry (PIV) often struggles in these zones because velocity signals are low and tracer patterns deform rapidly under high shear.
To address this, the authors developed PIV/Interface Gradiometry (PIV/IG+). This method utilizes covariance stacks and image transformations to directly estimate the velocity gradient (wall shear) at the boundary, even for complex curved surfaces. However, a critical operational question remains: how sensitive are these results to the precision of the wall's identified location? This paper quantifies the impact of sub-pixel errors in wall positioning on the accuracy of shear measurements.
Methodology
The PIV/IG+ method follows a structured six-step pipeline designed to extract gradients from raw imagery:
- Wall Identification: Defining the boundary manually or via algorithms.
- Conformal Transformation: Utilizing tools like the "SeaGrid" toolbox to map curved boundaries onto a rectangular grid, simplifying subsequent processing.
- Covariance Stacking: Applying 1D PIV to horizontal pixel lines to generate a "stack" of cross-correlation data.
- Peak Identification: Locating the strongest correlation signals within the stack.
- Curve Fitting: Fitting a straight line to the covariance peaks near the wall to determine the velocity gradient ($\partial U / \partial Y$). The vertical extent of data used for this fit is defined as the "fitting height" ($H$).
- Reverse Transformation: Mapping the results back to physical coordinates.
To test sensitivity, the researchers intentionally introduced shifts of $\pm 2$ pixels to the assumed wall position and observed the resulting variance in calculated wall shear.
Experimental Setup
The study utilized experimental data from a turbulent open-channel flow over a wavy bed. The physical model consisted of a 20 cm x 20 cm transparent PVC block with a CNC-machined sinusoidal upper surface (50 mm wavelength, 2.5 mm amplitude). Flow was captured using a double Nd:YAG laser and a Kodak ES1.0 camera at 30 fps, with the Matlab-based "SeaGrid" toolbox handling the orthogonal grid generation.
Key Insights: The "Lock-in" Effect and Sensitivity
A primary finding of the study is the "Lock-in" Effect. Covariance peaks located further from the wall (near the top of the template) tend to dominate the slope of the best-fit line. Because the fitting line is mathematically constrained to pass through the origin (the wall), any error in the wall's position forces the slope to pivot significantly to maintain its connection to these dominant peaks.
This leads to a critical trade-off regarding the fitting height ($H$): while a smaller $H$ provides a more localized measurement of the gradient, it drastically increases sensitivity to wall-positioning errors. The authors introduced a non-dimensional sensitivity parameter, $S$, finding that $S$ is consistently greater than 1.0. This indicates that the relative error in wall shear is actually larger than the relative error in the initial wall position.
Results and Conclusions
The investigation yielded several quantified benchmarks for PIV/IG+ accuracy:
- Gradient Sensitivity: Reducing the fitting height from 35 pixels to 15 pixels caused the measured average wall shear to increase by 50% for a fixed wall position.
- Quantified Error: The sensitivity parameter $S$ was approximately $1.25 \pm 0.10$ for mean wall shear and $1.35 \pm 0.15$ for RMS fluctuations.
- The Verdict on Precision: Because $S > 1$, the authors conclude that sub-pixel accuracy in locating the wall is mandatory for obtaining reliable wall shear measurements in turbulent flows.
Future work will focus on automating sub-pixel wall detection to mitigate human error and investigating how tracer density affects the minimum viable fitting height. While the sensitivity ranges identified here are consistent for this wavy-bed experiment, further research is required to generalize these findings across diverse flow conditions.
Summary updated