# A Softsensor for Wind Measurements in Karst Caves

## TL;DR

This paper introduces a data-driven soft sensor to measure wind velocity in caves as an alternative to physical anemometers.

## Contribution

A Gaussian process-based soft sensor is proposed for reliable wind measurements in karst caves.

## Key findings

- The soft sensor performs well based on statistical and visual evaluations of test data.
- Soft sensors can replace or supplement physical sensors in underground meteorology.
- They offer a cost-effective and reliable solution for filling data gaps.

## Abstract

A data-driven soft sensor of wind in a cave passage is developed as an alternative to physical anemometers for measuring wind velocity. It is intended to either fill data gaps during periods without physical measurements or to serve as a substitute for the physical sensor. It is implemented as a Gaussian process model, trained on one year of half-hourly measurements. Statistical measures and visual inspection of the test data indicate that both selected model structures perform well. Therefore, soft sensors represent a viable tool in underground meteorology. They may replace physical sensors that are fragile, power-intensive, or expensive. Alternatively, they can fill data gaps when a physical sensor is unavailable.