Parameter range reduction from partial data in systems of differential algebraic equations (Q2297179): Difference between revisions

An algorithm for parameter range reduction in systems of differential-algebraic equations (DAE) is presented. Given coarse initial ranges of the parameters and partial discrete time series data of observed outputs, the algorithm computes reduced parameter ranges and an approximation of each parameter value. To this end, interval arithmetic is applied to linear multistep discretizations of the DAE systems to discard parameter ranges that are inconsistent with the measured data. The algorithm is an extension of [\textit{A. Skelton} and \textit{A. R. Willms}, J. Sci. Comput. 62, No. 2, 517--531 (2015; Zbl 1320.65112)], where ordinary differential equations and measurements of all state variables where considered, to DAE systems and partial measurements. DAE systems of differential index 1 or higher have to be transformed to explicit ordinary differential equations; the original algebraic equations are nevertheless utilized by the algorithm. The authors apply the algorithm to three case studies the models of which have up to 5 state variables and up to 8 unknown parameter values. They demonstrate that parameter estimation methods, that use multistart local optimization, can benefit from the algorithm outputs.