Design Of An Intelligent Pumpover System For Wine Fermentation Process Based On Redox Potential

The fermentation process in the wine industry is highly mechanized,but the degree of automation and intelligence is relatively low,especially in the management of pumpover during red wine fermentation.Pumpover is a major task in red wine fermentation,usually requires manual operation of pumps and pipelines every day during the fermentation.This process is not only labor-intensive,but also poses significant safety risks for workers operating on top of large fermentation tanks.In this study,by analyzing the characteristic changes in redox potential during the fermentation process and combining machine learning techniques,an intelligent pumpover system based on fermentation potential was designed.The system makes decisions on pumpover management by monitoring the potential of red wine fermentation online,aiming to achieve unmanned production and ensure the quality of the final product to the greatest extent possible.The main research findings are as follows:1.An online monitoring system for redox potential was designed and implemented.The system utilizes online monitoring of the potential during red wine fermentation to analyze the relationship between potential changes and the behavior of pumpover,thereby achieving information-based pumpover.A Pt-Ag/AgCl electrode system was used as the potential sensor,installed in the fermenting must to collect the fermentation potential.The collected potential signals were converted into RS485 signals and sent to a Raspberry Pi,which handled data storage and remote transmission to the cloud.2.An automated pumpover system for red wine fermentation was designed and implemented.The system automates the pumpover operation,based on the relationship between potential changes and tank transfer behavior,controls fermentation potential by controlling pumpover.The Raspberry Pi’s GPIO pins were used to control relays,which in turn controlled the operation of pumps for automatic pumpover.The system also included fermentation tanks and a temperature control system designed for laboratory-scale fermentation.3.Based on the aforementioned online monitoring and automatic control systems,eight pumpover methods were designed for producing dry red wine.The physicochemical indicators of the eight wines were measured to analyze the quality of the finished products and determine the optimal tank transfer method for high-quality wine.Furthermore,the correlation between fermentation potential curve features and pumpover methods was established.This allowed the online monitoring and automatic control system to execute pumpover decisions based on the potential curve of the optimal pumpover method.The system autonomously determined whether pumpover was necessary during the red wine fermentation process,with a decision accuracy ranging from 97.38% to 98.04%.This achieved an intelligent pumpover system based on fermentation potential,further reducing manual labor intensity on top of the automatic pumpover system.Additionally,the system only made the error of prematurely stopping pumpover.4.Fermentation potential in wine fermentation is a time series data.When analyzing the relationship between fermentation potential and physicochemical indicators of the fermentation process,or the relationship between fermentation potentials,it is important to ensure the "two-dimensionality" of the time series data.Therefore,in this study,curve matching methods such as Fréchet distance,Global alignment kernel,Dynamic time warping,and Hausdorff distance were used to analyze the correlation of time series data.Furthermore,the classification algorithms used in the intelligent tank transfer system(Nearest neighbor,Support vector machine,Multilayer perceptron,and Shapelets)were called from time series data processing libraries to meet the analysis requirements of time series data.