It has been known that fuzzy system provides a framework to handle
uncertainties and vagueness, but the applications often face difficulties in deciding the number of inference rules, either in Sugeno or Mamdani fuzzy rules, and input/output membership functions. A five-layer neuro-fuzzy model is developed in this work and its model accuracy is validated by a nonlinear model benchmark test. Analysis shows that both Sugeno and
Mamdani neuro-fuzzy models have good performance in system identification, and the former achieves better accuracy.