Page 364 - Design of Simple and Robust Process Plants

9.2 Historical Developments 351
During the early 1990s, several complementary developments also took place:

. Computing power was increasing exponentially, and was no longer seen as a
constraint.
. Model-based controls such as predictive constraint control were applied at a
larger scale in the process industry, the refineries being the leaders in this
field.
. First applications of CLO based on NLP were introduced in refineries as well
as ethylene plants.
. The first commercial executive systems for CLO were installed.
. The optimizers could solve very large problems as LS-SQP (large-scale suc-
cessive quadratic programming)
During the 1990s, computing power developed so rapidly that it was no longer seen
as a constraint on OO, which has been implemented in several large chemical plants
next to olefin plants (Factora et al., 1992), at hydrogen, benzene, ammonia, nitric
acid, ethylene oxide plants.
The technology is still very young, and the practical applications are mainly lim-
ited to steady-state conditions. Future developments will include dynamic optimiza-
tion to be used for optimization of: transient operations, gradually fouling systems,
and dynamic optimization of continuous fed-batch reactor systems. Transient opera-
tions often taken place in continuous polymer plants that change from operational
conditions to produce other grades. The amount of transient material, which has a
lower product quality and value, can be minimized by the application of an optimal
change over. The optimization in fouling systems such as catalyst aging or coke for-
mation is another field for optimization. In these situations, the run length between
cleaning or regeneration and the operational conditions over the run length are sub-
ject to optimization. Dynamic optimization of batch reactor systems can increase
reactor capacity and product quality. The developments that are required in order to
implement these at commercial scale include:
. Accurate dynamic reactor and fouling models for the specific systems.
. Dynamic optimization techniques in a commercial format.
. Nonlinear model-based controllers.
. Accurate on-line measuring techniques for product properties with short
response times, particular optical techniques are full filling this opportunity.
. On-line model validation techniques.
These developments are already under progress within the academic world, but
need to be proven at plant scale.