Abstract:
Tuning methodology and other aspects of asimple fuzzy logic controller are studied
in this thesis. It is established that a zero order Sugeno type fuzzy controller is
analytically equivalent to a simple Mamdani type fuzzy controller. Comprehensive
analysis comparing the initial outputs of linear and proposed fuzzy controllers in the
context of derivative kick is presented. It is shown that simplest fuzzy controller if
properly tuned is capable of providing better performance as compared to an equivalent
well-tuned linear controller. Based on speed of response as criteria, three methods of
tuning are proposed. First two methods require the knowledge ofinitial outputs and gains
of a well-tuned linear controller for calculating the parameters of the equivalent fuzzy
controller. The third method is more conventional in approach and supports all the
conventional control design techniques. The design methodology, such as pole-zero
cancellation, damping accuracy, maximum overshoot, settling time, rise time criteria used
for linear control design can also be used for fuzzy logic based control design.
Simulation results comparing the performance of the fuzzy controller tuned using
proposed tuning methodology with equivalent linear and other ill-tuned fuzzy controllers
establish the superiority ofthe proposed tuning methodologies.
Robustness of proposed controller is studied for uncompensated second order process
with dead time against variation in the process dead time and process characteristics. The
use of Smith predictor compensator is proposed for making the fuzzy control system
insensitive to dead time variations. Simulation results for uncompensated and
compensated system clearly show that for larger errors between actual parameter values,
the performance ofthe proposed controller is much superior both from the point ofview
of response to reference input and unit step disturbance.
