Determination of Friction Characteristics using GRNN Embedded Model

R. Chancharoen (Thailand)


Friction, Artificial Neural Network, Fuzzy Inference System, System Identification.


This paper introduces a new scheme for determination of nonlinear characteristics of friction in a mechanical system, which contains unknown system parameters. The proposed technique models the entire system with GRNN Embedded Model, which combines the advantages of the Neuro-Fuzzy system with those of the Least–Squares Estimation. The upper level gradient-based learning algorithm is also presented. This algorithm manages the learning of the Neuro-Fuzzy and Least–Squares Estimation such that the model emulates the system. Simulation and experimental results demonstrate the advantages of the technique; when GRNN Embedded Model is closely fitted to the input/output data pairs, not only the nonlinear characteristic is obtained but the system parameter is also identified more accurately.

Important Links:

Go Back