For effective monitoring and control of the fermentation process, an accurate real-time measurement of important variables is necessary. These variables are very hard to measure in real-time due to constraints such as the time-varying, nonlinearity, strong coupling, and complex mechanism of the fermentation process. Constructing soft sensors with outstanding performance and robustness has become a core issue in industrial procedures. In this paper, a comprehensive review of existing data pre-processing approaches, variable selection methods, data-driven (black-box) soft-sensing modeling methods and optimization techniques was carried out. The data-driven methods used for the soft-sensing modeling such as support vector machine, multiple least square support vector machine, neural network, deep learning, fuzzy logic, probabilistic latent variable models are reviewed in detail. The optimization techniques used for the estimation of model parameters such as particle swarm optimization algorithm, ant colony optimization, artificial bee colony, cuckoo search algorithm, and genetic algorithm, are also discussed. A comprehensive analysis of various soft-sensing models is presented in tabular form which highlights the important methods used in the field of fermentation. More than 70 research publications on soft-sensing modeling methods for the estimation of variables have been examined and listed for quick reference. This review paper may be regarded as a useful source as a reference point for researchers to explore the opportunities for further enhancement in the field of soft-sensing modeling.

L-Lysine is produced by a complex non-linear fermentation process. A non-linear model predictive control (NMPC) scheme is proposed to control product concentration in real time for enhancing production. However, product concentration cannot be directly measured in real time. Least-square support vector machine (LSSVM) is used to predict product concentration in real time. Grey-Wolf Optimization (GWO) algorithm is used to optimize the key model parameters (penalty factor and kernel width) of LSSVM for increasing its prediction accuracy (GWO-LSSVM). The proposed optimal prediction model is used as a process model in the non-linear model predictive control to predict product concentration. GWO is also used to solve the non-convex optimization problem in non-linear model predictive control (GWO-NMPC) for calculating optimal future inputs. The proposed GWO-based prediction model (GWO-LSSVM) and non-linear model predictive control (GWO-NMPC) are compared with the Particle Swarm Optimization (PSO)-based prediction model (PSO-LSSVM) and non-linear model predictive control (PSO-NMPC) to validate their effectiveness. The comparative results show that the prediction accuracy, adaptability, real-time tracking ability, overall error and control precision of GWO-based predictive control is better compared to PSO-based predictive control.

The vital state variables in marine alkaline protease (MP) fermentation are difficult to measure in real-time online, hardly is the optimal control either. In this article, a dynamic soft sensor modeling method which combined just-in-time learning (JITL) technique and ensemble learning is proposed. First, the local weighted partial least squares algorithm (LWPLS) with JITL strategy is used as the basic modeling method. For further improving the prediction accuracy, the moving window (MW) is used to divide sub-dataset. Then the MW-LWPLS sub-model is built by selecting the diverse sub-datasets according to the cumulative similarity. Finally, stacking ensemble-learning method is utilized to fuse each MW-LWPLS sub-models. The proposed method is applied to predict the vital state variables in the MP fermentation process. The experiments and simulations results show that the prediction accuracy is better compared to other methods.

The marine protease fermentation process is a highly nonlinear, time-varying, multivariable, and strongly coupled complex biochemical reaction process. Due to the growth and reproduction of living organisms, the internal mechanism is very complicated. Some key variables (such as cell concentration, substrate concentration, and enzyme activity) that directly reflect the fermentation process's quality are difficult to measure in real-time by traditional measurement methods. A soft sensor model based on a support vector regression (SVR) is proposed in this paper to resolve this problem. To further improve the model's prediction accuracy, the grey wolf optimization (GWO) algorithm is used to optimize the critical parameters (kernel function width σ, penalty factor c, and insensitivity coefficient ε) of the SVR model. To study the influence of selecting auxiliary variables on soft sensor modeling, the successive projection algorithm (SPA) is used to determine the characteristic variables and compare them with grey relation analysis (GRA) algorithm. Finally, the Excel spreadsheet data was called by MATLAB programming, and the established SPA-GWO-SVR soft sensor model predicted crucial biological variables. The simulation results show that the SPA-GWO-SVR model has higher prediction accuracy and generalization ability than the traditional SPA-SVR model. The real-time monitoring was processed by MATLAB software for the marine protease fermentation process, which met the requirements of optimal control of the marine protease fermentation process.