# Automated monitoring of alcoholic fermentation: trends and challenges

## TL;DR

This paper reviews automated monitoring techniques for alcoholic fermentation, highlighting promising biosensors and the need for better data standards.

## Contribution

The paper provides a critical evaluation of current automated fermentation monitoring methods and identifies gaps in data science standards and public databases.

## Key findings

- Electronic nose and tongue biosensors show promise for monitoring alcoholic fermentation.
- Current automated monitoring studies lack reproducibility and industrial testing.
- Machine learning applications in these studies do not follow established data science standards.

## Abstract

The progress of fermentation, an important step in spirit production, needs to be monitored regularly to detect possible faults. Automated monitoring of fermentation, however, is often limited to only a few parameters of the mash such as, and mainly, its temperature. With the advance of sensor technology and data analytics, various solutions to automated fermentation monitoring emerged, mainly for the beer and wine industry, however, these are not yet critically evaluated and compared. Thus, scientific articles on automated monitoring of alcoholic fermentation are reviewed and evaluated here according to the type of sensors used, the type of fermented material, and the reproducibility and feasibility of the presented solutions. Possible data analytics methods to utilize are introduced and their pros and cons are discussed. A critical evaluation from scientific and industrial perspectives is provided with prospects for the distilling industry where mashes of various states of matter, inhomogeneity and viscosity can appear. Key findings and conclusions of this review are: Electronic nose and electronic tongue biosensors are a promising direction in the area. A publicly available database on recorded data from e-nose and e-tongue as well as other sensors on fermentation monitoring is needed but still missing. Current solutions on automated fermentation monitoring are rather isolated studies, conducted in laboratories, yet to be evaluated and tested in industrial environments. The use of machine learning techniques in these studies, in general, does not comply with the well-established standards in data science and artificial intelligence.