Page 363 - Design of Simple and Robust Process Plants

350 Chapter 9 Operation Optimization
9.2
Historical Developments

OO is an activity which was introduced gradually into the process industry. It began
in the 1960s and 1970s with the scheduling of batches and equipment for proces-
sing steps based on mixed integer linear programming (MILP).
Particular equipment selection optimization was developed and introduced at an
early stage for the large mechanical workshops, where the production was often lim-
ited by the different machines that were required for the different manufacturing
steps.
Based on LP techniques, production planning became also practiced. During that
same time-frame, off-line optimization for crude distillation at refineries started
based on LP with a Simplex optimization technique. The operation of the refineries
showed a higher productivity as a result of the implementation of the calculated
operational regimes for the different crude oils. These models were also applied to
feedstock evaluation. The introduction of LP optimizations in the process industry
started for scheduling within the typical batch processes of drug and food manufac-
ture.
At the beginning of the 1980s, commercial modeling software became available
commercially, based on non-linear programming (NLP)techniques with a sequen-
tial modeler for process simulation. The NLP techniques were very quickly accepted
as tools for process design, particular for the chemical industry. Most of these
designs are complex and non-linear. By the end of the 1980s, equation-based mode-
lers were available for process flowsheeting. The software was specifically developed
for much faster solution of very large problems, and also had the capability of hand-
ling dynamic simulation. The introduction of NLP equation-based modeling was
completed with an optimizer as SQP (Successive Quadratic Programming). This
was the first opening for operation optimization of non-linear continuous processes.
In the meantime, the capabilities of computing systems was passing through an
exponential growth curve.
The availability of these tools made it possible for them to be used initially for off-
line OO in the process industry. A number of off-line optimizations were developed
for those continuous process industries which were subject to varying markets and
feed stock conditions. It should be mentioned, that these variations should not be
too frequent (less than once a day). The models were also applicable for feed stock
evaluations. Refineries also converted their LP optimization into NLP optimizations,
to achieve a higher accuracy level, and this resulted in more savings and extended
the applications to other units such as hydro-crackers (Marlin and Hrymak, 1997).
A model with high accuracy makes it possible to operate closer to the plant con-
straints. Savings, reported by different companies, all showed that productivity
improvements were in the order of percentages of the operational cost for plants,
subject to variations. Engineers from process plants that were exposed to less vari-
able conditions learned that several plant conditions needed to be adapted signifi-
cantly to comply with optimal conditions, and so achieved considerable savings.