Set membership parameter estimation of fractional models based on bounded frequency domain data

Abstract

This paper deals with parameter estimation of fractional models based on frequency domain uncertain but bounded data. Fractional models parameters, including differentiation orders, are expressed as intervals. Real and complex interval analysis is then used to estimate the whole set of feasible parameters. Two methods based on interval constraints satisfaction problem (CSP) are compared to each other. The first one formulates a real-CSP based on an explicit decomposition of the frequency response in the real and the imaginary parts. The second one formulates a complex-CSP using the complex frequency response without any decomposition. To have a fair comparison, rectangular representation of complex intervals is considered. Contractors combined to bisection algorithms are used to solve both CSPs.