48. Chemometrics and Process Control

Measurement techniques for operando monitoring of catalytic chemical processes are reaching new levels of maturity. However, the primary outputs of monitoring devices, e.g., raw optical spectra, require multivariate interpretation or processing to extract useful information. Chemometric methods can achieve this in different ways, either in a mechanistic way using physicochemical principles or using statistical approaches. The concepts behind them are presented and compared, and the three most prominent methods – peak integration, spectral hard modeling, and projection to latent structures (PLS regression, also known as partial least squares) – are demonstrated in case studies using mid-infrared, Raman, and nuclear magnetic resonance (NMR) spectroscopy. The underlying principles are transferrable to other techniques as well if certain constraints remain fulfilled. The workflow style of the chapter is expected to efficiently assist users which are new to the field in solving their challenges in multivariate spectral analysis.