Abstract

In situ local void fraction in the vertical air-gas two-phase flow has been detected by the combination of multiple current-voltage (MCV) and plural long short term memory lasso (MCV-pLSTM-L) for the flow status monitoring in cooling equipment of nuclear reactor. The combination of MCV and machine learning has been applied to void fraction measurement of vertical air-gas two-phase flow. However, in our previous research Plural Long Short Term Memory Lasso, the machine learning employed drift flux model to calculate the true void fraction. There was a possibility of disagreement between true in situ void fraction and calculated void fraction. MCV-pLSTM-L has five steps, 1) true local void fraction measurement by electrical probe sensor, 2) voltage measurement by the MCV system, 3) voltage extraction by Lasso, 4) flow regime identification by 1st LSTM, and 5) void fraction α estimation by 2nd LSTM. Experiments were conducted in a vertical pipe with an inner diameter of 25mm and a length oh 80mm from the inlet. Two vertical flow regimes which were liquid single-phase flow and bubbly flow were used. As a result, the local void fraction is successfully estimated and applicability of MCV-pLSTM-L is confirmed.

This content is only available via PDF.
You do not currently have access to this content.