| Summary: | 碩士 === 大同工學院 === 電機工程研究所 === 81 === In this thesis, a neural network fuzzy modeling/control system
which is designed by the combination of the neural network
control system and the fuzzy logic control system is proposed.
This proposed system combines the learning ability of neural
network and the advantage of fuzzy logic controller to handle
the nonlinear system modeling and control problem. First, a
fuzzy modeling method using neural network with the back
propagation algorithm is presented. Then, this architecture is
trained to control nonlinear plants and back up a simulated
truck to a loading dock in a parking lot. As the learning
ability of neural network control system, the proposed system
can be trained by experience data to find the optimal number of
fuzzy logic rules and the optimal parameters of input/output
membership functions. This feature is important particularly as
the knowledges of human expert are unavailable. Next in the
design of the neural-network fuzzy modeling/control system, the
architecture of neural-network control system can be simplified
to accelerate the process of self-learning by the use of fuzzy
logic control rules. From the simulation results, the proposed
scheme can be employed to approximate a nonlinear function very
well by self-learning from training data. Regardless of the
number of inputs and outputs of the plant under consideration,
one can get the optimal number of fuzzy logic rules and the
optimal membership functions. Furthermore, the neural-network
fuzzy modeling/control system can model/control most of the
complex ill-defined system without knowing the mathematical
models of them.
|
|---|
