Intelligent Tuning of Fuzzy Controllers by Learning and Optimization - Fuzzy Modeling and Control: Theory and Applications - page 130

this chapter, and it is based on the classification of soft-computing components, first
proposed by Bonissone et al. ( 1999 ), Bonissone ( 2000 ) and updated in a later work
by Verdegay et al. ( 2008 ).
In particular, this chapter is focused on the use of learning and gradient-free
optimization techniques for optimal tuning. Two case studies are also reported in
order to demonstrate the suitability of the proposed strategies.

7.2 Fuzzy Systems in Single Loop and Internal
Model Control Schemes

There are several types of systems that use a fuzzy logic device (FLD) as their main
system component, and therefore many schemes for modeling and control systems
are available. Indeed, the right choice of architecture contributes to the fulfillment
of the system's objectives. Some years ago, Wang and Tyan ( 1994 ) have suggested a
classification of the most useful schemes for fuzzy logic-based modeling and fuzzy
control:

• FLD for input selection (type 1) (Zadeh 1988 ),

• FLD for feedback error/output control (type 2) (King and Mamdani 1977 ),

• FLD for controling the parameters of dynamic systems (type 3) (Kosko 1991 ),

•

FLD for choosing the best compensator (type 4) (Kim et al. 1996 ),

•

FLD derived from a multiple performance index (type 5) (Miyamoto et al. 1987 ),

• FLD as a mathematical model for unknown or complex system dynamics (type 6)

(Narendra et al. 1995 ),

• Hierarchical FLD (type 7) (Lin et al. 1997 ).

Indeed, we must select the appropriate control and modeling schemes for accom-
plishing our goal. In this chapter we will focus on single loop feedback error and
internal model control schemes. Control applications based on fuzzy and neurofuzzy
systems require a design methodology that must satisfy control requirements. The
internal model control (IMC) paradigm accomplishes this goal by using direct and
inverse process models for designing the control system (Morari and Zafiriou 1989 ).
From a conventional viewpoint, the internal model control method uses a closed-
loop control scheme containing both a direct model ( G M ) parallel to the process to be
controlled and an inverse model ( G M ) preceding the process ( G P ). Disturbances are
represented by d . Figure 7.2 shows the a single loop control system and the internal
model control schemes.
The use of the IMC paradigm theoretically guarantees control system robustness
and stability in the presence of external disturbances. However, non-linear model
inversion is not an easy task. Analytical solutions may not exist; thus, solutions have
to be calculated numerically. In addition the inversion of the process model may lead
to unstable controllers when the system has a non-minimum phase.