Development of a recursive time series model for fed-batch mammalian cell culture

Abstract

Recursive time series models are developed in this work for a fed-batch mammalian cell culture producing monoclonal antibodies, with key culture variables measured at different sampling frequencies. Glucose and glutamine feed rates are considered as inputs. A composite of an autoregressive moving average with exogenous input model and a dual rate-autoregressive with exogenous input model is used. Appropriate parameter constraints are imposed in parameter estimation algorithms and stability of these is examined and ensured. The data required for parameter estimation are generated from simulated fed-batch experiments using a well-tested first principles model. The predictions for glucose, glutamine, and viable cell concentrations track very well the data for these, with the errors for the high prediction horizons considered being limited to 10% or less. The prediction accuracy can be increased further if data from prior experiments with dynamic similarities are available. The models can be used reliably for model predictive control.