Odour sensing in turbulent plumes with high-speed electronic nose and non-invasive ground truth

TL;DR

This paper introduces a high-resolution dataset combining optical and electronic nose measurements of turbulent odour plumes, enabling improved sensor modeling and reconstruction in real-world environments.

Contribution

It provides a benchmark dataset with synchronized optical and electronic nose data for assessing and developing odour sensing algorithms in turbulent conditions.

Findings

High-speed electronic nose responses can be quantitatively analyzed using the dataset.

The dataset supports benchmarking of reconstruction and deconvolution algorithms.

Open data and scripts facilitate research in odour sensing and related fields.

Abstract

Chemical sensing in real-world environments requires resolving rapidly fluctuating and spatially heterogeneous concentration fields. However, these dynamics are strongly distorted by widely used, low-cost metal-oxide (MOx) gas sensors, whose thermal and surface-kinetic response acts as a low-pass filter on the underlying concentration signal. Quantifying and compensating for these effects remains challenging, largely due to the lack of benchmark datasets that simultaneously capture the spatiotemporal structure of turbulent odour fields and the time-resolved response of point sensors. Here, we present a dataset combining planar laser-induced fluorescence (PLIF) measurements of an acetone tracer plume with synchronised recordings from a custom, kilohertz-rate microelectromechanical (MEMS) MOx electronic nose deployed in a laboratory wind tunnel. The PLIF system provides quantitative, two-dimensional concentration fields at high spatial and temporal resolution, while the co-located e-nose records film resistance, heater currents, and environmental parameters with aligned timestamps. The dataset enables quantitative assessment of sensor dynamics, development and benchmarking of reconstruction and deconvolution algorithms, and data-driven modelling of plume structure. All recordings, metadata, calibration files, and example analysis scripts are released in open, platform-independent formats. Together, these provide a valuable reference for researchers working in odour-guided robotics, environmental monitoring, computational fluid dynamics, and neuromorphic sensing, supporting the design and evaluation of high-speed odour-sensing systems.