NARMAX model of the pCO/sub 2/ control system in man estimated by neural computation

Abstract

Subjects voluntarily inspire a gas mixture in which the CO/sub 2/ concentration is changed stepwise or randomly. The respiratory flow rate and pCO/sub 2/ in the inspired and expired gases are measured to yield the end-tidal pCO/sub 2/ and minute ventilation, which are the input and output of the pCO/sub 2/ control system, respectively. A NARMAX (Nonlinear Auto-Regressive Moving Average with eXogeneous inputs) model of the system is estimated using a three-layered feedforward neural network. The estimated model contains terms, y(t-1), x(t-1), x(t-2), X/sup 2/(t-2) and y(t-1)x(t-2). A measure of nonlinearity calculated from the data used for estimation shows the pCO/sub 2/ control system in most subjects has a nonlinearity which cannot be neglected.