Résumé:
In this work, two approaches for designing fuzzy inference systems from data are
developed. These approaches are characterized by their ability to automatically extract and
improve knowledge from numerical data and to preserve interpretability of fuzzy rules during
the optimization process.
In the first approach, a Neuro-Fuzzy Controller network, called NFC that implements
a Mamdani fuzzy inference system is proposed. This network includes neurons able to
perform fundamental fuzzy operations. Connections between neurons are weighted through
binary and real weights. Then a new algorithm called mixed Binary-Real Non dominated
Sorting Genetic Algorithm II (MBR-NSGA II) is developed to perform both accuracy and
interpretability of the NFC by minimizing two objective functions. In order to preserve
interpretability of fuzzy rules during the optimization process, some constraints are imposed.
The approach is tested on two control examples: the pole and cart system and a helicopter
simulator model.
In the second approach, a genetic algorithm based method for designing fuzzy wavelet
neural network (FWNN) is presented. The proposed framework combines several soft
computing techniques such as fuzzy inference system, wavelet neural network and genetic
algorithm. Thus, the structure of the proposed FWNN consists of combination of two network
structures, one containing the fuzzy reasoning mechanism and the other containing Wavelet
neural networks. Then a genetic algorithm based method is used to find optimal values of the
parameters of the both network structures. The ability of the technique in identifying non
linear dynamical systems is demonstrated on two examples. Also, this approach is tested for
the control of two dynamic plants commonly used in the literature.
