Page 335 - Design of Simple and Robust Process Plants

8.4 Control Design 321
DMC), and fuzzy logic control (for an overview, see Qin and Badgewell, 1997). All
these techniques require the development of models through process identification
(also called input and output models). Compared to fundamental, model-based con-
trollers, these are handicapped by their accuracy. They can be totally out of order in
case new conditions occur, such as other feed streams for which the models were
not trained, or in case insufficient data are available to achieve a model of sufficient
accuracy. However, they have as an ultimate advantage the tenet that detailed pro-
cess knowledge is not required. The development of hybrid models was expected by
Qin et al. (1997). Hybrid models integrate steady-state, non-linear, first-principles
models with dynamic empirical models. The ultimate solution is based on the devel-
opment of dynamic models which undoubtedly will evolve. The disadvantage of
dynamic simulation as a time effort will disappear. Dynamic simulations will
become easier to program, while object-oriented programming and robust library
models will become available. Long execution times of the simulations will com-
pletely disappear as computer power is increased progressively. Another argument
which plays a role is that modeling technology leads to improved process knowl-
edge. The process knowledge required is very helpful to design more integrated pro-
cesses which have to meet increased demands on safety and environmental require-
ments, which as such must result in robust control.
The objective of this section is to describe a methodology for robust control design
of the basic control layer, based on fundamental models ± static as well as dynamic.
A fundamental approach is preferred, as in the long term this will be the only struc-
tural way to achieve consistent robust control designs. Nevertheless, historical
results have been achieved using an heuristic approach.

8.4.1
Control Strategy Design at Basic Control Level

The basic control layer is dominated by feed-back loops, eventually provided with
feed-forward actions to cope with disturbances. The basic control layer will include
some cascaded control loops and simple models, such as for de-coupling of interac-
tions, and heat and mass balance control options. The demand on control, next to
hands-off control, is higher than in the past due to:
. more stringent requirements on product quality;
. elimination of intermediate storage as lot tanks, check tanks/hoppers and
minimization of storage, (implementation of JIP, TQC, and FPPP);
. high level of process integration;
. switchability of process conditions for campaign operations (switchability is
defined as the ease with which the process can be moved from one stationary
point to another);
. predictive alarming and interlocking actions need to be incorporated to
assure safe operations (see Section 8.3.3);