SENSE -- Sensor-Enhanced Neural Shear Stress Estimation for Quantitative Oilfilm Visualizations

TL;DR

SENSE introduces a neural network-based method that combines oil-film optical flow analysis with sparse sensor data to accurately quantify wall shear stress in fluid dynamics experiments, overcoming noise and resolution limitations.

Contribution

The paper presents a novel neural network framework that integrates sparse sensor measurements with oil-film optical flow to enhance shear stress estimation accuracy.

Findings

Over 30% reduction in root-mean-squared error compared to classical methods.

Robustness to sequence length and spatial resolution variations.

Effective global regularization from sparse sensor data improves estimates far from sensors.

Abstract

Wall shear stress quantification is fundamental in fluid dynamics but remains challenging in wind-tunnel experiments. Sensor-based methods offer high accuracy but lack spatial resolution for capturing complex three-dimensional effects. Conversely, oil-film visualization is a simple method to obtain high-resolution surface flow topology by processing a sequence of images using optical flow (OF) techniques. However, leveraging this approach for quantitative analysis suffers from noise and systematic biases. This study introduces SENSE (Sensor-Enhanced Neural Shear Stress Estimation), a data-driven approach that leverages a neural network to enhance OF-based shear stress estimation through the integration of sparse, high-fidelity sensor measurements via a multi-objective loss function. SENSE processes oil-film image sequences directly, inherently mitigating temporal noise without explicit averaging. The method is validated in a turbulent separated flow on a one-sided diffuser. Results demonstrate SENSE's robustness to sequence length and spatial resolution compared to classical optical flow algorithms. Crucially, incorporating sparse sensor data significantly improves quantitative accuracy, achieving over 30% reduction in root-mean-squared error on validation sensors with only 8 strategically distributed sensors. The sensor data provides a global regularization effect, improving estimates far from sensor locations. SENSE offers a promising approach to elevate oil-film visualization to a reliable quantitative measurement technique by combining image sequences and sparse sensor data.