(371k) Multi-Rate Data-Driven Models for Lactic Acid Fermentation - Parameter Identification and Prediction

Recursive time series models are developed in this article for bacterial cultures producing lactic acid from lactose. Biomass concentration, X, is measured much less frequently than are lactose and lactic acid concentrations. A composite of an autoregressive (AR) model and a dual rate-autoregressive with exogenous input (ARX) model is used. Appropriate parameter constraints are imposed in parameter estimation algorithms and stability of these is examined and ensured, enabling reliable output prediction. The models for X allow for prediction of X at the frequency of faster sampled concentrations with inter-sample output estimation, a feature useful in process optimization and model predictive control. The data required for parameter estimation are generated from simulated batch experiments using well-tested first principles models for two bacterial strains. The predictions for lactose, lactic acid, and biomass concentrations track very well the data for these. The prediction accuracy can be increased using data from prior experiments.