Chemical Modelling - Sponsored Whitepaper

Improving prediction robustness from hard and soft modeling

Jplus Consulting was formed in 2009 to develop and commercialise innovative multivariate data analysis tools for the quantitative characterization of chemical processes. The ReactLab™ suite of software provides sophisticated but easy to use applications for analysing both process and equilibrium titration spectroscopic measurements and yields all the underlying reaction rates, equilibrium constants as well as the spectra and concentration distributionsoftheparticipatingchemicals pecies.

Working with an industry leader Dr Peter King, Managing Director of Jplus Consulting, says “Very early on we started a dialog with CAMO Software who recognized the complementary nature of our `hard modeling' approaches to the classical chemometric multivariate analyses such as PCA and PLS and `soft modeling'techniquesincludedinTheUnscramb ler®X .” Essentially both approaches analyse chemical data to gain information about the underlying process or system using mathematical and statistical tools. The key difference is that the chemometric algorithms do not require a starting model but allow the development of an abstract model from patterns and behavior in training samples which can then be used to predict the constitution and key characteristics of unknown samples. These characteristics may not be `real' parameters such as absorption spectra or concentration profiles but abstract mathematical decompositions, which nonetheless allow very powerful prediction of unknowns. “The Unscrambler® X is very well equipped in this area and offers a wide range of toolsinanexceptionallyuserfriendlyenviro nment”saysPeter.

Rigorous mathematical model for chemical processes `Hard modeling' starts with a mathematically rigorous model of the process under investigation. In the case of ReactLab, this is the reaction scheme underlying a chemical process. This model is then fitted to the measurement. This involves the adjustment and optimisation of the key rate and equilibrium parameters using a least squares approach. This yields the fundamental reaction parameters that define the process which can be re-used in morecomplexsystemsandbyotherresearchers. ReactLab™ is an ideal tool for studying dynamic chemical systems, either kinetics, where reactions are monitored spectroscopically as a function of time in order to determine rates and yields of a process, or chemical equilibria, where the data are collected during titration experiments in order to quantify ligand binding interactions or complexation equilibria. “We're looking at specific cases with CAMO where the marriage of these two approaches can be brought bear on process monitoring applications involving robust characterization of the process in the lab, coupled with online spectroscopic measurement of the subsequent process. By applying a combination of ReactLab™ & The Unscrambler® X methods we aim to provide robustonlinepredictionsofcomponentdistri butionsandprocessquality”explainsPeter.

Industry: 4Multivariate data analysis software vendor