Nonlinear identification of the PCO2 control system in man

Two approaches to identification of the PCO2 system in man are described. The first uses a nonlinear 'black box' NARMAX identification package, while the second method uses a structured two-compartment Belville model. The data were obtained from volunteers breathing either room air or a controlled gas mixture, controlled via a pseudorandom M-sequence. Measurements were made of respiratory gas flow and PCO2 content of inspired and expired gases. The identification results indicate that a low-order dynamic model with nonlinear polynomial expansion gave the best fit to the data. In contrast, the Belville model gave best results with a two-compartment linear model, mainly because of difficulties in the optimisation routines when the Belville model was not linear. Thus, modern systemic methods of excitation and identification appear to be appropriate for modelling this respiratory subsystem of humans.