Real-time monitoring
Real-time monitoring of critical quality attributes (CQA) or critical
process parameters (CPP) using real-time monitoring by means of soft
sensors (Mandenius et al., 2015), (Roch et al., 2016) has enormous
advantages over the determination of those off-line after the unit
operation. It is much faster and can assess attributes, where no
dedicated sensor or analytical method is available. It also provides
information in real-time if the process is within the operating range.
These soft sensors are also called prediction models (e.g., (Sauer et
al., 2019)). Often, we also want to monitor other attributes, which are
beyond the CQA, because they are relevant for process economics, but not
relevant in respect to patient safety. In a conventional process,
particularly in DSP the samples are taken from a pool or intermediates
and then subjected to further off-line analysis. Interventions in the
upstream process do not require a fast reaction because cells grow
slowly and a response after up to an hour is sufficient. Even an at-line
measurement using automatic sampling devices with a 30-min lag time is
considered as a “real-time” measurement because the return of the
measurement is fast enough to control the process (Zhao et al., 2015).
Whereas in a lot of situations in downstream processes the decision must
be made within seconds, e.g., when a stream is diverted from waste to
product collection. Then the decision must be made within seconds and
time-consuming analytics become useless.