Generalized dynamic modeling of local heat transfer in bubble columns

Instantaneous local heat transfer rates were measured by using a hot-wire probe in three bubble columns of different diameters of 200, 400 and 800 mm. The time series of heat transfer rates were analyzed by means of rescaled range (R/S) and deterministic chaos analyses. Due to the influence of highly chaotic bubble motions, the instantaneous local heat transfer exhibits low-dimensional chaotic features. The dependences of Hurst exponents and Kolmogorov entropy on the column scale consistently suggest different nonlinear hydrodynamic behaviors exist in bubble columns of different scales. Based on the measurement of instantaneous heat transfer rates, an artificial neural network (ANN) was applied to correlate instantaneous local heat transfer with dynamic motions of bubble and liquid. The ANN was optimized and trained by only using the experimental data measured at one location of 200 mm column. The trained ANN model shows good performance for the generalized use to predict the dynamic heat transfer rate in three columns over whole experimental conditions studied, indicating the ANN is capable of capturing the universal relation between instantaneous heat transfer and local bubble dynamics.