Detection of the chaotic flow instability in a natural convection loop using the recurrence plot analysis and the nonlinear prediction

An experiment and a time series analysis have been made on the flow in a rectangular natural convection loop. It was found that the flow reversal occurs in a certain range of the heat flux and Reynolds number, which depends on the cooling water temperature. Based on time series temperature data of the system, an attractor in a state space has been reconstructed. Geometrical structures and orbital instability for this attractor were evaluated by the recurrence plot analysis and a nonlinear prediction by the method of local linear approximation. A nonlinear prediction was applied to the turbulence time series. As a result, a recurrence plot of the predicted time series obtained from the turbulence data has characteristics similar to those of the measured of the unstable flow. This fact suggests that, even when an apparent turbulent flow is observed, a transition to the unstable flow may occur. In order to confirm this conjecture, we have performed another experiment with the addition of a disturbance to the loop. The experimental results are consistent with the chaos analysis. We have successfully shown the validity of the nonlinear prediction and its recurrence plot as a means to evaluate dynamical flow instability.