Chemometrics for on‐line spectroscopy applications—theory and practice

In today's chemical industry there is a great need to both improve the effectiveness of existing assets and generate new prospects for business renewal. In addition, the fast pace of the current ‘information age’ makes it necessary to bring about both of these in a rapid and responsive manner in order to be competitive. On‐line spectroscopic techniques, which provide rapid and representative information on a chemical process, are inherently capable of facilitating both asset productivity and business renewal. Chemometric ‘tools’, in principle, are particularly well suited for improving the effectiveness of on‐line spectroscopic techniques through two main functions: (1) extracting the wealth of useful, yet convoluted information that is available from such techniques; and (2) facilitating the automation of on‐line analytical techniques, at a time when manufacturing plant support resources are being ‘cut back’. Multivariate calibration tools can be used to improve the performance of existing methods, as well as to prepare more stable methods for new applications. More ‘soft’ modeling tools, such as multivariate curve resolution, are particularly well suited for providing timely information during the early stages of new process development, where very little is known about the process chemistry and dynamics. In practice, however, there is a wide range of issues, both technical and non‐technical, that limit and even prevent the successful application of chemometric techniques for on‐line spectroscopy. This paper will illustrate both the means by which chemometric tools can make a difference in the chemical industry, and the issues that are preventing its successful implementation, using examples from actual process development projects and on‐line measurements. Copyright © 2000 John Wiley & Sons, Ltd.