FCANN: An Approach to Knowledge Representation From ANN Through FCA Effects of Synthetic Data Base and Discretization Process, Application in the Cold Rolling Process

Abstract

Nowadays, artificial neural networks (ANN) are been widely used in the representation of physical process. Once trained, the nets are capable to solve unprecedented situations, keeping tolerable errors in their outputs. However, humans cannot assimilate the knowledge kept by those networks, since such knowledge is implicitly represented by their connection weights. Formal concept analysis (FCA) can be used in order to facilitate the extraction, representation and understanding of rules described by ANN. In this work, the approach FCANN to extract rules via FCA will be applied to the cold rolling process. The approach has a sequence of steps as the use of a synthetic database and intervals of discretization where the data number variation per parameter and the intervals variation of discretization is an adjustment factor to obtain more representative and precision rules. The approach can be used to understand the relationship among the process parameters through implication rules.