Flow regime identification of co-current downward two-phase flow with neural network approach

Hiroshi Goda, Ye Mi, Joshua P. Finch, Seungjin Kim, Mamoru Ishii, Jennifer Uhle

    Research output: Contribution to conferencePaperpeer-review

    10 Scopus citations


    Flow regime identification for an adiabatic vertical cocurrent downward air-water two-phase flow in the 25.4 mm ID and the 50.8 mm ID round tubes was performed by employing an impedance void meter coupled with the neural network classification approach. This approach minimizes the subjective judgment in determining the flow regimes. The signals obtained by an impedance void meter were applied to train the self-organizing neural network to categorize these impedance signals into a certain number of groups. The characteristic parameters set into the neural network classification included the mean, standard deviation and skewness of impedance signals in the present experiment. The classification categories adopted in the present investigation were four widely accepted flow regimes, viz. bubbly, slug, churn-turbulent, and annular flows. These four flow regimes were recognized based upon the conventional flow visualization approach by a high-speed motion analyzer. The resulting flow regime maps classified by the neural network were compared with the results obtained through the flow visualization method, and consequently the efficiency of the neural network classification for flow regime identification was demonstrated.

    Original languageEnglish (US)
    Number of pages8
    StatePublished - Oct 19 2002
    Event10th International Conference on Nuclear Engineering (ICONE 10) - Arlington, VA, United States
    Duration: Apr 14 2002Apr 18 2002


    Other10th International Conference on Nuclear Engineering (ICONE 10)
    Country/TerritoryUnited States
    CityArlington, VA

    All Science Journal Classification (ASJC) codes

    • Nuclear Energy and Engineering


    Dive into the research topics of 'Flow regime identification of co-current downward two-phase flow with neural network approach'. Together they form a unique fingerprint.

    Cite this