Fourier transform (FT)-Raman combined with partial least squares regression (PLS-R) calibration models allows the accurate monitoring of solids content, copolymer composition, and free amounts of monomers in starved semi-batch emulsion copolymerizations. The calibration models remain valid as long as the spectrometer and the measuring conditions are unchanged. Unfortunately, maintenance and/or repairing of the spectrometer result in changes in the relative intensities of the peaks of the Raman spectrum, reducing the performance of the calibration models. Therefore, a strategy for the up-date of the PLS-R calibration models is needed. Strategies for calibration model maintenance were assessed, and we found that the best strategy was to build a new model composed of the old PLS-R model plus a PLS-R model able to account for the model mismatch of the old model.