Next generation combined sonic-hotfilm anemometer: wind alignment and automated calibration procedure using deep learning

Abstract

The study of naturally occurring turbulent flows requires ability to collect empirical data down to the fine scales. While hotwire anemometry offers such ability, the open field studies are uncommon due to the cumbersome calibration procedure and operational requirements of hotwire anemometry, e.g., constant ambient properties and steady flow conditions. The combo probe-the combined sonichotfilm anemometer-developed and tested over the last decade has demonstrated its ability to overcome this hurdle. The older generation had a limited wind alignment range of 120 degrees and the in-situ calibration procedure was human decision based. This study presents the next generation of the combo probe design, and the new fully automated in-situ calibration procedure implementing deep learning. The elegant new design now enables measurements of the incoming wind flow in a 360-degree range. The improved calibration procedure is shown to have the robustness necessary for operation in everchanging open field flow and environmental conditions. This is especially useful with diurnally changing environments or non-stationary measuring stations, i.e., probes placed on moving platforms like boats, drones, and weather balloons. Together, the updated design and the new calibration procedure, allow for continuous field measurements with minimal to no human interaction, enabling near real-time monitoring of fine-scale turbulent fluctuations. Integration of these probes will contribute toward generation of a large pool of field data to be collected to unravel the intricacies of all scales of turbulent flows occurring in natural setups.

Atmospheric Small-Scale Turbulence from Three-Dimensional Hot-film Data

“…13, varying from 0.3 to 0.7. These values are relatively high but of the same order of magnitude of the values obtained by Kit and Liberzon (2016) and Goldshmid et al (2022) using both traditional and neural network calibration. We expect that in a more favorable setup, such as sonic pointing to the streamwise direction, the sonic velocity would correspond to a better "ground truth" for the velocity fluctuation and the δ i values would be lower.…”

“…This is a cause for concern, since scaling and isotropy are frequently used as measure of the reliability of the data. A more carefully designed field experiment is needed to address some of these questions, ideally including a pre-calibration of the hot-film probe to check all calibration parameters, and a method for reorientation of the sensors in the field (as already present in the combo probe by Goldshmid et al (2022), for example). Future work could also include a validation of the neural network calibration method in the small-scale statistics of the flow.…”

“…14. Furthermore, we estimated the delta parameter as a quantitative measurement of the difference between the two time series (Kit and Liberzon 2016;Goldshmid et al 2022). The delta parameter is defined as…”

“…When employed in the Mountain Terrain Atmospheric Modeling and Observations (MATERHORN) field experiment, Kit et al (2017Kit et al ( , 2021 used a combo probe calibrated with a neural network to investigate turbulent bursts and structure functions, respectively, within a 90min period of stably stratified flow. Recent developments updated the combo probe calibration into a deep learning neural network approach (Goldshmid et al 2022), eliminating the human-decision-based selection of data for the neural-network training, and improving its automatization.…”

Bubble Detection with Semantic Segmentation for Multiphase Flow Particle Image Velocimetry

“…The authors found that the deblurred images resulted in less PIV velocity error by applying the generator to actual experimental images of a synthetic jet. Authors of [8] presented a study concerning a combined sonic-hotfilm anemometer, and a fully automated in situ calibration procedure implementing DL. The calibration procedure was tested on both the open sea dataset and inside a wind tunnel, it has shown to have the robustness necessary for operation in everchanging open field flow and environmental conditions.…”