Neural computing in pharmaceutical products and process development - Computer-Aided Applications in Pharmaceutical Technology - page 160

Proceeding of a successful granulation in SOM
through regions A (mixing phase), B (spraying
phase), and C (drying phase). Gray areas
represent the undesirable regions of the map
(Ia-Ib, blocking of fi lters; IIa-IIc, low level of
relative humidity of inlet air) (reprinted from
Rantanen et al., 2001; with permission of
Elsevier)

Figure 5.19

￿

￿ ￿

optimal path through the SOMs and to identify undesirable regions
(i.e. undesirable processing conditions) (Figure 5.19).
Since SOMs give information about correlation between different
processing and control properties, they can be used to summarize
multiple process control charts and diagrams and for determination
of correction actions.
In complex processes, computer-aided process monitoring and
fault diagnosis enhances the operator's abilities to run the process
smoothly. A toolbox was developed to monitor the process state
using Kohonen SOMs in conjunction with heuristic rules (Jämsä-
Jounela et al., 2003).