Page 334 - Design of Simple and Robust Process Plants

320 Chapter 8 Instrumentation, Automation of Operation and Control
ªThe ease with which a process can be controlled to achieve a desired perfor-
mance within specified limits determined by capacity, product quality, and environ-
mental and safety constraintsº (Bogle, 1989).
The developments to support hands-off control has undergone major progress
during recent years, and this is reflected by the large number books on control and
control design (e.g. Liptak, 1995; Marlin and Hrymak, 1997; Luyben et al., 1998; Sko-
gestad and Postlethwaite, 1996). It is also reflected in the attention given by Seider
et al. (1999) in his book Process Design Principles, which include chapters on plant
wide controllability assessment.
The design of control hierarchy requires a layered approach ± also called a decen-
tralized approach ± as was illustrated in Figures 8.1 and 8.2: Wolff et al 1992. This
concept, which is generally accepted as the basis for a robust operated process,
demands a very robust basic control layer which can be easily handled by operation,
in case the model-based and optimization layer are not functioning properly and are
switched off. The operation needs to proceed hands-off, at some appropriate dis-
tance from its constraints and its ultimate optimum operational point. The model-
based control layer is designed for multi-variable controllers to de-couple interaction
and to approach constraints and support optimization of operation. The inputs and
outputs always pass the basic control layer, where the output are set points for basic
control loops to ensure independent operation of the basic control layer.
Major progress has been made in the design of a robust basic control layer, which
found a basis in the following developments:

. Availability of robust, equation-based dynamic simulators, with extended
library models for unit operations.
. Development of a method for plant wide control (Luyben et al., 1998).
. The design methodology for selecting the dominant control and manipulated
variables based on a thermodynamic approach (TyrØus 1999).
. The design methodology for self-optimizing control which minimizes inter-
action by selecting appropriate controlled variables based on economics
terms (Skogestad et al., 1999).
. Control strategy selection based on static and dynamic interaction parameters
used for selection.
. Implementation of selected control strategies in a simulation for testing
robustness and control algorithms on disturbance rejection.
. Development of de-coupling algorithms at basic control layer preferable by
algebraic equations (Verwijs, 2001).
The technology developed from heuristic design methods to more fundamental,
model-based design methods. These developments still based on linear controllers
are the cornerstones for the design of a robust basic control layer.
During the past few decades, most of the control developments have concentrated
on the development of multi-variable, model-based controllers. These developments
were focused on the de-coupling of interactions and model predictive constraints
control. Several types of model-based controllers were developed, including neural
net controllers, dynamic matrix controllers (DMC; later extended to Quadratic