Abstract:

A new method was developed for realtime monitoring the Nisin titer (NT) and the concentration of cell during Lactococcus lactis subsp. Fermentation by near infrared spectroscopy (NIR) combined with radial basis function neural network (RBFNN). Three different 5 L fermentors were applied to implement 15 batches of the Lactococcus lactis subsp. Fermentation. Samples were collected interval 1 h during these fermentations. The Nisin titer and the cell concentration were determined by reference methods. At the same time the NIR spectra of the samples were recorded using UVVisNIR spectrophotometer. RBFNN was applied to model the relationship between NIR spectra and Nisin titer and the cell concentration. The RBFNN models were optimized by selecting efficacious preprocessing methods, wavelength, the number of hidden nodes and the spread constant. The optimum models for determination of the NT and the cell concentration were developed. The coefficient of NIR predictive values and reference values of calibration set were 0.8649 and 0.9914, the root mean square error of prediction set (RMSEP) were 2865.05 and 0.1414,respectively. These results demonstrate that fit and the predictive capability of these models were satisfied. This method can be used for monitoring the key parameters during fermentation processes.

Key words: near infrared spectroscopy, radial basis function neural network, Nisin titer, Lactococcus lactis subsp fermentation