Soft sensor of biological parameters in the marine protease fermentation process

In order to solve the difficulties of online measurement in the marine protease fermentation process of crucial biological variables (such as biomass concentration, substrate concentration and enzyme activity, etc), a soft sensing method based on KPCA-RBF neural network is proposed by combining the kernel principal component analysis (KPCA ) with the radial basis function (RBF) neural network. Establishing the soft sensing model of KPCA-RBF neural network, KPCA is applied to compress data, and choose the nonlinear component as the input of RBF neural network and biomass concentration, substrate concentration, relative enzyme activity as the output. Simulation results indicate that this model has a higher accuracy, better tracking performance when compared with RBF and PCA-RBF neural network model. Therefore, the proposed method can satisfy the requirements of on-line measurement of biological parameters and is proved to be an efficient modeling method.