For conventional PID tuner and fuzzy inference systems based on expertise problem will arise when expertise of the process is not enough. Artificial neural networks have self-learning capability, however, the change of their weights can not be understood, This paper describes the structures of self-learning neuro-fuzzy networks and shrinking-span membership functions, and presents a neuro-fuzzy PID (NFPID) controller. The NFPID controller has the capability of self-extracting inference rules, and its parameters have explicitly physical definitions. By using the RBF neural network inverse model, a hybrid learning procedure was put forward. Various simulation results demonstrated that the NFPID controller described has very good performances.
[1]
Jyh-Shing Roger Jang,et al.
ANFIS: adaptive-network-based fuzzy inference system
,
1993,
IEEE Trans. Syst. Man Cybern..
[2]
J. G. Ziegler,et al.
Optimum Settings for Automatic Controllers
,
1942,
Journal of Fluids Engineering.
[3]
C.-L. Chen,et al.
Design and analysis of neural/fuzzy variable structural PID control systems
,
1996
.
[4]
Chuen-Tsai Sun,et al.
Functional equivalence between radial basis function networks and fuzzy inference systems
,
1993,
IEEE Trans. Neural Networks.
[5]
C.-L. Chen,et al.
User-friendly design method for fuzzy logic controller
,
1996
.