Modified Soft-sensing Method of Erythromycin Fermentation Process Variables Based on "Assumed Inherent Sensor" Inversion

In this paper, we proposed a soft-sensing method of erythromycin fermentation process variables based on "assumed inherent sensor" inversion, which is modified from the method we previously proposed. The main feature of the modified method is the utilization of the non-state variables or so-called function variables, which will greatly increase the possibility in constructing the soft-sensing model and make the soft-sensing model easy to implement. Furthermore, with the algorithm steps reformed, the method ensures the prior utilization of the lower order derivatives, which will decrease the sensitivity to the noise and will be of great benefit to the engineering application.