Biomedical Engineering Reference
In-Depth Information
Figure
12.9. Online real-time detection and diagnosis of a process deviation enables engineers
to troubleshoot the process more effectively. Panel
(a)
score time series plots,
(b)
T
2
time series
plots,
(c)
batch maturity plots,
(d)
DModX (or squared prediction error) plots, and
(e)
executive
batch dashboard for two consecutive bioreactors (first two plots or boxes in the dashboards for
two phases in the first bioreactor, batch and fed-batch). (See the insert for color representation
of this figure.)
Note that the dashboard shown in Fig. 12.9e also indicates (color codes) the
fed-batch phase of this particular batch as out-of-multivariable control. Therefore in
essence, one can simply monitor the dashboard to get the executive summary of process
in-control state in real time and focus on the deviation by interactively exploring the area
of focus that is needed due to any deviation. This hierarchical approach not only
simplifies engineers and operators' supervision tasks but also makes themmore efficient
in a proactive manner.
Besides usingmultivariate approaches, a number of soft-computing techniques have
also been developed to monitor biopharmaceutical manufacturing process performance.
Examples include using the existing measurements to derive calculated variables
or statistics to achieve increased sensitivity in monitoring performance and obtain
predictions. In addition, practical approaches are proposed to overcome data alignment
problems when measurements are made only with respect to time. One common
approach includes monitoring the progress of a batch with respect to an indicator
variable that can also be considered as relevant time variable so that the rest of the process
variables can bemonitored against its progression points over time. This process variable
is selected to indicate the progress of a batch instead of time. Each new measurement is
then sampled relative to the progress of this variable. This variable, which indicates the
