Digital twin for predicting and controlling food fermentation: A case study of kombucha fermentation

Abstract

In the era of rapid advancements in computing and the Internet of Things, the food fermentation sector is undergoing a digital and intelligent transformation. This research developed a food fermentation prediction and control system based on digital twin technology. The system employs multi-scale feature extraction and convolution feature fusion to establish partial least squares (PLS) prediction models for C source and bacterial concentration. The results showed that the PLS prediction models of C source and bacterial concentration exhibited excellent performance, with RMSEP of 0.5538 mg/mL and 0.0558 (Au), and RPD of 5.63 and 6.52, respectively. An optimal control system for the fermentation process was constructed by integrating the prediction models with a genetic algorithm (GA), yielding satisfactory simulation and testing outcomes. The study showed that the proposed digital twin-based fermentation prediction and control system offers superior robustness and reliability, advancing the digital and intelligent development of the food fermentation industry.