Soft sensor modeling based on GD-FNN for microbial fermentation process

In order to solve the problems in real-time measurement of crucial biological parameters (such as biomass concentration, substrate concentration, product concentration, etc.) in the microbial fermentation process, a soft sensor modeling method based on Generalized Dynamic Fuzzy Neural Network (GD-FNN) is proposed. Taking the penicillin fermentation process as an example, initially, the auxiliary variables of the soft sensor model are determined by means of the uniform incidence degree method. Secondly, generation criteria of fuzzy rules that contain ε-completeness are ascertained by use of the elliptical basis function. Finally, the soft sensor model is established by GD-FNN. Simulation results show that soft sensor modeling based on GD-FNN is faster in operating speed, and has a higher forecast precision and better generalization ability than Radial Basis Function (RBF) neural network.