Diagnosing model inadequacy in human respiratory CO/sub 2/ controller

Ventilatory noise in human CO/sub 2/ controller identification is considered, and a test procedure that helps to provide a solution to it is proposed. Noise sources are examined and a new procedure developed to test model inadequacy. The procedure is capable of checking the nature of ventilatory noise and its interaction with chemical drive, as well as detecting any possible nonlinearity in the controller. A table of test results of residuals with different noise and system models is presented and has been used to guide controller identification. Test results of human experimental data suggest that (a) the linear assumption of the controller is valid, (b) noise is additive, and (c) noise is either white or colored.<>