Multiple-sensor predictive chemometrics approach
If only a single quality attribute is monitored such soft sensors are often based on single sensors, e.g., the monitoring of the product quantity by UV spectroscopy (Rolinger et al., 2021) or attenuated total reflection Fourier-transform infrared (ATR‐FTIR) spectroscopy (Rudt et al., 2019). However, multiple sensors must be combined even if only the process signals are used that a chromatographic workstation is typically equipped with, e.g., UV, conductivity and pH probe (Nikita et al., 2022). If several CQAs need to be modelled in real-time, it is indispensable to use a multiple sensor approach in order to capture the different properties of these CQAs (Sauer et al., 2019). Such an approach can be regarded as data fusion (Borras et al., 2015; Liggins et al., 2017; Rolinger et al., 2021). A multiple sensor approach has many advantages, e.g., in the case of sensor saturation where the combination of different input variables leads to redundancies that can compensate for the information loss.