(196e) Hybrid Niche Algorithm for Problem Inversion of Distributed Systems with Solution Multiplicity
Many important chemical reaction systems have strongly coupling in heat and mass transfer. These systems are described with partial differential equations (PDEs) and method of the transport. We have introduced a kinetic inversion problem (TKIP) method to estimate the parameters for these distributed models. However, the inversion problems are characterized multiple solutions and many conventional numerical methods fail to find all solutions. In order to determine which solution constitutes the actual physical response or which parameter set best explains the actual measurements, we propose to identify first all locally optimum solutions. In the second stage, we revisit all local minima to select most physically meaningful outcome. Therefore it is necessary to deploy inversion algorithms which we can reliably identify all local minima for distributed inversion problems. For this reason, we developed ?a hybrid niche algorithm', based on the sequential niche technique (Beasley et al 93) in combination with deterministic problem inversion. In this method, we combined a genetic search with a deterministic method to accelerate locating multiple solutions to the parameter inversion problem. The method uses a self-adjusting niche radius to overcome the problem of clustering in the search space. Thus every niche area has a different size and updated. In this presentation, we will show the efficiency and robustness of our algorithms with several examples.