Non-invasive on-line two-phase flow regime identification

A novel non-invasive methodology is proposed for the classification of BWR two-phase flow regimes. Neutron radiography images (frames) of coolant flow recordings (videos) are used. Following image pre-processing via directly computable statistical operators, the extracted statistical characteristics are input into an ensemble of self-organizing maps. The individual decisions are averaged, resulting in the accurate on-line classification of each image into the corresponding flow regime