Application of a kernel method in modelling friction dynamics
| Authors: | Harrison Robert, The University of Sheffield, United Kingdom
Wan Yufeng, The University of Sheffield, United Kingdom
Wong Chian X., The University of Sheffield, United Kingdom
Dodd Tony J., The University of Sheffield, United Kingdom |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Nonlinear System Identification - Kernel Methods |
|---|
| Keywords: | Kernel method, Volterra series, NARX models, friction dynamics,neural network, system identification |
|---|
Abstract
A kernel method has been developed to model finite degree, finitememory length and infinite degree, finite memory length Volterra series usingpolynomial and exponential kernels, respectively. Here, the kernel method isextended to identify NARX (Nonlinear AutoRegressive model with eXogenousinputs) models. To verify its effectiveness, the proposed approach is used inmodeling friction dynamics, which is an important and complex mechanicalprocess. The simulation results are compared with those from a physical modeland a neural network, and the advantages and disadvantages of the methods areshown.
