Page 412 - Design of Simple and Robust Process Plants

9.6 Project Methodology for Operation Optimization 399
Ongoing operation attention looks simple, but it is not unusual that some appar-
ently minor modification is not noticed (at least from the perspective of the OO).
One might think about modifications in piping, equipment location, constraints,
control configuration or tuning (less steady state)or feedstock selection. A solution
might be to appoint the responsibility to a professional for most large process plants
such as refineries, and indeed this is common practice.
Instrumentation maintenance must concentrate on signal analysis. If instruments
are selected and installed according to the latest practices, calibration is no longer
required and element reliability is very high and maintenance is less of an issue.
Analyzers ask for a higher level of ongoing maintenance, although automatic checks
and calibrations are now available and are a major step forward, although sampling
and injection systems as applied for GCs are maintenance intensive.
Control demands ongoing attention to assure that control loops remain closed
and stable. The principle that operation should not touch control actions is a precon-
dition also for BC to maintain correct functioning. Any modification in the plant
might ask for updating of the control design, and should be seen as part of the
project.
Model updating is an effort like that of model-based control design, and requires a
high-tech detailed process and modeling knowledge. The models require updating
for any change in the plant even it is only a temporary one, as it always impacts on
the OO.
In the long term, an efficient maintenance structure is crucial to ongoing benefit
from OO.

9.6.15
Other Types ofOptimization

These include off-line optimization, mixed integer optimization, and dynamic opti-
mization.
. Off-line optimization is often the first step before the successive steps to qual-
ity and constraint control, and ultimately CLO are taken (see below for more
details). The model needs to meet a reasonable level of accuracy, and also
needs a validation effort. Therefore, a performance meter is a requirement,
but the process must also be sufficiently stable to achieve a reasonable valida-
tion, and estimation of the parameter which will be fixed in this model. The
model, however, will in general describe the constraints not accurately but
only in a very rudimentary form. Off-line optimization is mostly done if the
frequency of disturbances in the economic environment is low, but it might
be complemented with self-optimizing control.
. Mixed integer optimization (MILP and MINLP). Scheduling of operations is
generally done in MILP. Also, blending operation as practiced in refineries
and executed under fully automated operation are also mostly done in MILP.
These MILP scheduling and operations activities are beyond the scope of this
work. MINLP are performed for utility supplies such as power and steam.