Nonlinear Characterization of Pressure Fluctuations in Fluidized Beds

Nonlinear time series analysis techniques were applied to predict pressure fluctuation data in fluidized beds in two different hydrodynamic states. The method of delays was used to reconstruct the state space attractor to carry out analysis in the reconstructed state space. The state space reconstruction parameters, i.e., time delay and embedding dimension, were determined and the results shown that their values were different for various types of methods introduced in the literature. Chaotic behavior and predictability of fluidized system were determined by introducing two nonlinear dynamic invariants, correlation dimension and entropy, in different ways. The traditional linear autoregression method and state space based prediction methods (SSBPMs), i.e., nearest neighbors and locally linear, and global linear methods, were applied to predict the pressure fluctuation signals. The quality of prediction was assessed by comparison of the predicted data with its original benchmark. In addition, the dynamic invariants of measured and predicted attractor of the pressure signals were compared. The results showed that SSBPMs are preferred to the traditional linear methods. Finally, a continuous uncertainty band of pressure signals of single and multiple bubble regimes for the prediction methods was presented.