(76f) Process Monitoring Using Robust Chemometric Spectrum Models for Predicting Concentration Profiles

Heterogeneous catalytic processes are of extreme importance to the entire range of chemical industries, as the reactions that are being carried out increase in complexity, so to must the tools used to analyze and interpret the reaction data. Robust and accurate chemometrics/spectrocopy methods allow rapidly determination of the concentrations within reactors operating under industrial reaction conditions thereby enabling on-line optimization of such reaction conditions to maximize the desired product and reduce both energy costs and waste. The papers listed below discuss some of the techniques used for in situ monitoring on bioprocess, which are particularly sensitive to reaction environments and therefore usually have tighter controls.

This is the reason why research in chemical kinetics in addition to the measurement of reaction rates also involves the study of reaction intermediates by computation, spectroscopy, isotopic labeling, etc. Once reliable data for key intermediates become available these data constraints the reaction mechanisms and thereby the predicted macroscopic kinetics. This approach enables the discrimination between different reaction mechanisms. The availability of a robust PLS model provides a technique by which an estimation of the local concentrations from spectroscopic data, i.e. Raman, IR or ATR intensities, can be obtained at a very short timescale. The obtained concentration estimates of each reacting species can be subsequently used to recursively update the reaction rate coefficients in a given model, which typically change with temperature for example.

For more complex systems such as those often encountered in heterogeneously catalyzed reactions, the performance of the catalyst may also change with temperature or the reaction may switch to a different mechanism and hence, the structure of the defining rate equations can also vary. Using a total of three sets of experimental data, a Raman spectroscopy and a calibration experiment as well as recorded data from a debutanizer unit, the contribution of the presented work are two detailed application study which involve (i) recorded data from a Raman spectroscopy experiment and (ii) data from a calibration experiement. These applications highlight the merits and requirements of a robust PLS parameter estimation and show the practical usefulness of newly developed diagnosis charts to determine which Type a specific outlier is assoicated with. The presentation finally shows how to utilize these PLS models to identify abnormal process behavior through the on-line analysis of concentration estimates.

References

Lynggaard, H., Andreasen, A., Stegelmann, C., Stoltze, P., ?Analysis of simple kinetic models in heterogeneous catalysis?, Progress in Surface Science, Vol. 77, No. 3-4, pp. 71-137, 2004.

Vojinović, V., Cabral, J.M.S., Fonseca, L.P., ?Real-time bioprocess monitoring Part I: In situ sensors?, Sensors and Actuators B: Chemical, Vol. 114, No. 2, pp. 1083-1091, 2006.