Intelligent Tuning of Fuzzy Controllers by Learning and Optimization - Fuzzy Modeling and Control: Theory and Applications - page 131

Fig. 7.2 Two of the most widely applied control schemes: single loop feedback control and internal
model control
The filter G F is included in the control system in order to reduce the high-frequency
gain and improve system robustness. It also helps soften sudden and abrupt signal
changes, thereby improving controller response.
An IMC scheme can be implemented using a neurofuzzy system. First the neu-
rofuzzy system is trained using real-world input/output data so that it learns the
process dynamics, and a direct process model is obtained. Another neurofuzzy sys-
tem is trained to learn the inverse process dynamics and performs as a non-linear
controller, obtaining the inverse model. For the direct model, feed rate is taken as
the input variable and mean cutting force as the output variable. In order to adjust
the model, a set of training data is used to create an initial neurofuzzy model. Then,
a second set of data is used to adjust the initial system parameters.
7.3 Intelligent Tuning of Fuzzy Control System
In general, one of the main shortcomings when designing a control system is the need
of apriori knowledge of the structure of the controller (data base, rule base, decision-
making and defuzzification) before an approximation is made. Unfortunately, in
most real cases, it is not easy to define the structure or 'functional form of the fuzzy
controller, and any decision in this regard can subjectively influence by the nature of
the problem. Figure 7.3 shows some key aspects and relevant parameters that can be
considered for designing and tuning a fuzzy control system.
Since the middle of 90s the scientific community focused the attention on the use
of soft-computing techniques to aid an initial fuzzy controller design and, in some
cases, to setup a fuzzy controller from a very limited knowledge.