A Fuzzy Sensitivity Analysis for Improved Parameter Identification in Human Metabolism Models

A. Kistner, M. Hanss, and O. Nehls (Germany)


Fuzzy arithmetic, sensitivity analysis, parameter identifi cation, human metabolism models


Identifying a greater number of parameters in a complex process model may become a difficult task when the set of available measurement data from the process is limited and when it contains data only from some of the relevant process variables. In such cases standard identification procedures may run into convergence problems. The situation gets better as soon as some at least vague ideas about the ranges of the parameter values and about their presumed distributions within these intervals are available. In the contribution a new refined parameter identification technique will be presented which can be applied then. Herein the uncertain parameters are modeled as appropriate fuzzy numbers. Using a special transfor mation method, an efficient implementation of a fuzzy arithmetic may be formulated and applied for analyzing the influences of each of the uncertain parameters to the variations of the overall model outputs. The resulting sen sitivity information can be used to distinguish between various phases in the process dynamics and to form corre sponding subsets of parameters which easily can be iden tified on the basis of measured data from one or the other of these phases and those which cannot. The benefits of the technique will be shown with a com plex physiological model for the glucose metabolism of patients suffering from insulin dependent diabetes melli tus.

Important Links:

Go Back