ABSTRACT
In-line near-IR moisture monitoring of the dynamic, fluid bed drying environment has been reported in recent years by several research groups; however, analytical figures of merit with regard to prediction accuracy are discussed in only a subset of this work, and issues with sampling and sample presentation are scarcely addressed at all. In this study, experiments were performed at 65-, 300-, and 600-L drying scales using several different sampling configurations in an effort to better understand and improve in-line near-IR method accuracy. Findings from this work demonstrate that process heterogeneity plays a major role in determination of apparent prediction accuracy. This aspect is general to all in-line measurements and plays an especially important role in solids and slurry systems that are prone to heterogeneity. In addition to experimental results, simulations based on these findings and sampling theory demonstrate an interesting paradox: depending on the sampling configuration employed, the method with the smallest apparent error is not necessarily optimal for process monitoring and control. Furthermore, sampling configuration influences the number of samples necessary to define an adequate calibration set. Finally, process understanding that was gained as a result of temporally rich, in-line measurements will be presented.
