A prediction model of the fermentation process for rhamnolipid production under temperature control

Abstract:

Background Rhamnolipids are widely used biosurfactants, while their production is limited by various factors, which make it difficult to meet the demands of large-scale production. Additionally, due to the high costs of raw materials and the complexity of the fermentation process, the production cost remains high.Objective Taking temperature as the core variable, we explored the influences of temperature on microbial growth and generation of metabolic products during the fermentation for rhamnolipid production. The aim is to optimize the fermentation conditions, reduce production costs, and increase the rhamnolipid yield, thus providing a theoretical basis for industrial production.Methods By constructing a GRU-attention-KAN neural network model, we predicted the fermentation conditions for rhamnolipid production at different temperatures.Results The GRU-attention-KAN model demonstrated good prediction performance across three experiments, with the best group achieving a root mean square error of 0.022 8, a mean absolute error of 0.119 2, an R² value of 0.825 9, and an explained variance of 0.920 6.Conclusion The model effectively captures the dynamic effects of temperature changes on the fermentation process, exhibiting high prediction accuracy and application potential, thus laying a foundation for subsequent optimization efforts.