Quality prediction of a fermentation process using multi-scale GRU with hybrid modeling strategy

Fermentation is a crucial process in the production of food, pharmaceuticals, and other products, but its inherent complexity often makes it challenging to consistently meet quality standards. This paper proposes a multi-scale gated recurrent unit-based quality prediction method using a hybrid modeling strategy to enhance product quality consistency. This hybrid approach integrates the advantages of regression models in extracting precise and effective features with the superior performance of classification models in handling complex decision boundaries. Specifically, the regression network maps process variables to quality-related features, while the classification network then assigns these features to the corresponding product quality categories. Additionally, changes in the metabolic demands of microorganisms and the continuous consumption of substrates give rise to multi-stage dynamics, thereby further complicating the modeling process. To address this issue, a multi-stage segmentation approach is implemented within the classification network to more effectively capture the distinct phases of fermentation. Moreover, stages with minimal impact on training performance are excluded, streamlining and accelerating the model training process. Experimental results demonstrate that the proposed approach significantly improves the accuracy of quality predictions, thereby achieving more consistent quality control in fermentation processes.