| Summary: | 碩士 === 逢甲大學 === 自動控制工程所 === 91 === The purpose of this thesis is to evaluate the possibility of neural network linearization in identification and control. For the system identification, the linearized transfer function is more comprehensive and analyzable than the original neural network model; Further, it can be applied to system control design. Most of industrial PID controller design will depend on experiences of engineer and trial and error
approach to tune controller parameters. In this paper, first, we try to construct the architecture of the PID parameter-learning network. Secondly, the capability of the
auto tuning in neural network is adopted to accomplish the tracking control for the
reference model.
|
|---|
