Numerical Analysis of Model Parameter Uncertainties as a Result of Experimental Uncertainty — An Example from Preparative Chromatography

Abstract Model calibration, and in particular model parameter uncertainty caused by experimental errors, is the focus of this work. Computer simulations were used to design a purification step for insulin by reversed-phase chromatography. The effect of errors in the protein sample concentration and purity, and in the modifier concentration in the sample, equilibration, and elution buffers was studied on the calibration of the adsorption kinetic parameters by the inverse method. The overall error, including experimental errors, was not normally distributed and not uncorrelated. Monte Carlo simulations were performed where the calibrated model was used to generate new data sets and a random error was added on the experimental conditions. New model parameter sets were found by recalibrating the model to the data sets from Monte Carlo simulations and the model parameter covariances were estimated from these. A control strategy which was robust to uncertainty in both model and process was designed from the resulting model parameter distribution and the expected variations in the process variables.