Upward vertical two-phase flow local flow regime identification using neural network techniques

Traditionally, the flow regimes in two-phase flow are considered in a global sense. However, a local flow regime is required to understand and model the interfacial structures present in the flow. In this work, a new approach has been used to identify both global and local flow regimes in a two-phase upward flow in a 50.8 mm internal diameter pipe under adiabatic conditions. In the present method, the bubble chord length distributions, which are measured simultaneously with three double-sensor conductivity probes, have been used to feed a self-organized neural network. The global flow regime identification results show a reasonable agreement with the visual observation for all the flow conditions. Nonetheless, only the local flow regimes measured at the center of the pipe agree with the global ones. The local flow regime combinations found are analyzed using the flow map information, cross-correlations between the probe signals, and previous correlations. In this way, it is possible to identify eight different global flow regime configurations.