Development and application of an ANN-perception-based autonomous control system for Escherichia coli cultivation process

To address the challenges of overflow metabolism and the heavy reliance on manual intervention in high-density Escherichia coli fermentation, this study introduces an AI-driven, autonomous intelligent control system. Using superfolder green fluorescent protein (sfGFP) as a reporter, the research first optimized DO-stat feeding parameters and the induction process, achieving a 52.85% increase in cellular specific fluorescence intensity and significantly enhancing protein expression levels. Subsequently, an artificial neural network (ANN) model was developed and trained to achieve real-time recognition of dissolved oxygen (DO) baselines ( R 2 = 0.998). This model was integrated with feeding control logic to form the NeuroStat-Ctrl system, enabling fully autonomous control across the entire fermentation lifecycle. Utilizing this system, unattended E. coli fermentation was successfully achieved, with fluorescent protein production further increasing by 5.87% compared to the optimized manual control. Experimental validation demonstrated that the system effectively mitigates feeding deviations inherent in traditional fixed-threshold strategies, prevents metabolic overflow, and enhances process stability and reproducibility. Furthermore, this system provides an efficient, standardized, and intelligent solution for high-throughput strain screening and process validation in parallel bioreactors.