The invention relates to the technical field of
microorganism co-culture
fermentation, and discloses a method for producing L-
carnitine soybean meal by applying a neural network to optimize
fermentation conditions, and the method comprises the following steps: establishing a double-bacterium single-culture control group, collecting a double-bacterium dual-
wavelength response
data set, and constructing a two-dimensional geometric feature space to obtain a strain feature
fingerprint slope; establishing a double-bacterium co-culture
fermentation group, decomposing the real-time mixed
optical density value to obtain double-bacterium estimated concentration, and constructing an enhanced input
data set; then building a bimodal prediction model to output a predicted value of the L-
carnitine content; and finally, setting an optimization objective function containing a yield deviation item, a phase regression item and a control
smoothing item, and solving to obtain an
optimal control input sequence. According to the invention, precise decoupling of the growth state of double
bacteria and metabolic phase cooperative regulation and control are realized, information dimension limitation of traditional mixed
optical density signals is broken through, and
time sequence matching and yield stability of L-
carnitine synthesis are guaranteed.