Case Study on the Identification of a Direction-Dependent Electronic Nose System

Abstract

The identification of an electronic nose system having direction-dependent properties, where the dynamics depend on whether the output is increasing or decreasing, is described. The system and experimental setup are explained and results from step response tests are presented where the difference in dynamics can be observed. It is shown that the use of maximum length binary signals allows the detection of the nonlinearities through the input–output crosscorrelation function due to the shift-and-multiply property. When inverse-repeat maximum length binary signals are applied, the effects of even-order nonlinearities can be eliminated. The detection of the even-order nonlinearities is also possible through analysis of the output spectrum. The best linear approximation of the system is estimated using various methods in the time and frequency domains. Finally, it is shown that the identification tests lead to the estimation of a Wiener model for the system. The application of the inverse-repeat signal is advantageous here, as the odd-order and even-order components can be optimised separately resulting in higher accuracy in the parameter estimates.