Please use this identifier to cite or link to this item: https://doi.org/10.2316/P.2012.785-025
DC FieldValue
dc.titleAn application of narx-network to reduce contour errors
dc.contributor.authorHuo, F.
dc.contributor.authorChen, P.C.-Y.
dc.contributor.authorPoo, A.-N.
dc.date.accessioned2014-06-19T05:31:44Z
dc.date.available2014-06-19T05:31:44Z
dc.date.issued2012
dc.identifier.citationHuo, F.,Chen, P.C.-Y.,Poo, A.-N. (2012). An application of narx-network to reduce contour errors. Proceedings of the IASTED International Conference on Engineering and Applied Science, EAS 2012 : 209-215. ScholarBank@NUS Repository. <a href="https://doi.org/10.2316/P.2012.785-025" target="_blank">https://doi.org/10.2316/P.2012.785-025</a>
dc.identifier.isbn9780889869523
dc.identifier.urihttp://scholarbank.nus.edu.sg/handle/10635/73151
dc.description.abstractThe problem of improving the contour error in two-dimensional CNC machines is considered in this paper. The nonlinear autoregressive with exogenous inputs (NARX) network is a dynamic neural architecture commonly used for input-output modeling of nonlinear dynamic systems. In the work presented here, two sets of off-line trained NARX networks are used to predict the position outputs at the next sampling time instant for the two axes of a CNC machine. From these values, the expected axial components of the contour error for the next instant is computed and used to adjust the reference position inputs to compensate for this error. The inputs to the NARX networks are the original uncompensated reference position inputs and actual axial positions together with corresponding values in past instances, the number of these latter depending upon the complexity of the dynamics of the system. An iterative procedure is used to improve compensation performance. Simulation results using linear, circular and parabolic contours show that this approach can significantly improve contouring accuracy. Although model-based in its control strategy, this approach does not require an accurate knowledge of the system's dynamic model as the NARX networks are trained using actual input-output data which can be readily obtained from the system during operation.
dc.description.urihttp://libproxy1.nus.edu.sg/login?url=http://dx.doi.org/10.2316/P.2012.785-025
dc.sourceScopus
dc.subjectContour error
dc.subjectModel prediction
dc.subjectNARX neural network
dc.typeConference Paper
dc.contributor.departmentMECHANICAL ENGINEERING
dc.description.doi10.2316/P.2012.785-025
dc.description.sourcetitleProceedings of the IASTED International Conference on Engineering and Applied Science, EAS 2012
dc.description.page209-215
dc.identifier.isiutNOT_IN_WOS
Appears in Collections:Staff Publications

Show simple item record
Files in This Item:
There are no files associated with this item.

Page view(s)

240
checked on Sep 23, 2021

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.