Software platform for the statistical validation of structural partitioning methods

Abstract

A novel interactive software platform is proposed to support design and development for PSE. It has been implemented to address the statistical validation of structural partitioning methods. According to several features given as model parameters, it produces incidence matrices whose structural partitioning can lead to a more efficient resolution of such models. The global objective is to generate automatically an arbitrary number of incidence matrices, shaped on the basis of statistical parameters associated with real-world PSE models. Then, partitioning methods can be executed on the generated matrices. Computational results for several problem instances are reported. Realistic cases were chosen by increasing the model-complexity level: a standard distillation column and an ammonia synthesis plant. In particular, the Direct Method, the Extended Direct Method and the Improved Extended Direct Method (IEDM) were evaluated. In comparison, the IEDM exhibited statistically significant enhancements of efficiency values for the resolution of the corresponding models.