Neural Network and Chaotic Characteristics Analysis of  Water Film on Corrugated Plate Wall

The corrugated plate dryer is an important steam-water separation device in marine nuclear power systems. The flow characteristics of the free falling liquid film on the corrugated board wall have a great effect on the steam-water separation efficiency of the dryer and the economic and safety indicators of the marine nuclear power plant. The liquid film thickness of different Reynolds number is measured by the Plane Laser Induced Fluorescence (PLIF) technique. Based on the small data volume method, the maximum Lyapunov exponents of liquid film thickness time series under different experimental conditions are calculated. The chaotic characteristics of the liquid film on the corrugated plate wall and phase space reconstruction is analyzed and performed, respectively. The BP (back propagation) neural network which has the advantage of solving nonlinear problems is applied to predict the liquid film thickness, the prediction model of single hidden layer BP neural network is established, and the nonlinear characteristics of the thickness of the free liquid film is analyzed. Results reveal that the maximum Lyapunov exponent is positively correlated with the liquid film Reynolds number. The liquid film volatility which is increased by the isolated peaks generated in the large Reynolds number region are coupled with gravity and the superposition effect of different liquid films, which makes the chaotic characteristics of the liquid film more obvious.