malgorzata sumislawska prof keith j burnham coventry university
DESCRIPTION
Parity equations-based unknown input reconstruction for Hammerstein-Wiener systems in errors-in-variables framework. Malgorzata Sumislawska Prof Keith J Burnham Coventry University. Motivation. Errors-in-variables (EIV) framework - PowerPoint PPT PresentationTRANSCRIPT
Univ logo
Parity equations-based unknown input reconstruction for Hammerstein-Wiener
systems in errors-in-variables framework
Malgorzata SumislawskaProf Keith J Burnham
Coventry University
UKACC PhD Presentation Showcase
Univ logo UKACC PhD Presentation Showcase Slide 2
Motivation Errors-in-variables (EIV) framework
Input and output signals are subjected to white, Gaussian, zero-mean, mutually uncorrelated measurement noise sequences
Long history of research on EIV framework in Control Theory and Applications Centre
Aim: reconstruct unknown input while minimising impact ----of measurement noise on unknown input estimate
Univ logo UKACC PhD Presentation Showcase Slide 3
Motivation Hammerstein-Wiener (HW) system representation considered
Relatively simple model structure Can approximate large class of nonlinear systems Limited attention paid to HW systems in EIV framework
N1(.) , N2(.) – static nonlinear functions
Univ logo UKACC PhD Presentation Showcase Slide 4
Problem solution Knowing N1(.) and N2(.) calculate input and output to linear
dynamic block Input and output estimates to linear block affected by noise
signals to be calculated
Univ logo UKACC PhD Presentation Showcase Slide 5
Problem solution Knowing N1(.) and N2(.) calculate input and output to linear
dynamic block Input and output estimates to linear block affected by noise Linear EIV setup with heteroscedastic noise, whose variance
depends on operating point Need for adaptive scheme
Univ logo UKACC PhD Presentation Showcase Slide 6
Problem solution Influence of noise minimised using Lagrange multipliers
optimisation method Time-varying noise variances estimated from N1(.) and N2(.)
using Taylor expansion Experimental (Monte-Carlo simulation) results match
theoretical calculations
Univ logo UKACC PhD Presentation Showcase Slide 7
Summary and future work Summary
Novel approach for unknown input reconstruction developed Effect of measurement noise minimised in adaptive manner The work published in Sumislawska M., Larkowski, T., Burnham, K. J., ‘Unknown input
reconstruction observer for Hammerstein-Wiener systems in the errors-in-variables', Proceedings of 16st IFAC Symposium on System Identification, Brussels, Belgium, 11-13 July 2012
Future work Coloured output noise Multivariable case