Application Of Artificial Neural Network And Chaos Theory In Gas-Solid Circulating Fluidized Bed

Gas-solid circulating fluidized bed is an high efficient equipment in mass-transfer, heat-transfer and gas-solid reaction. It has been used more and more widely in modern industry. The complex fluid flow and transport mechanisms make control, prediction and scale-up of gas-circulating fluidization process very difficult. In this paper, we apply artificial neural network and chaos theory to study the system for promoting the ability of its application and understanding the nonlinear hydrodynamics in the riser.All experiments were carried out in a circulating fluidized bed with high riser. Compared with the experimental set-ups used by many other research groups, this riser was higher, and therefore the gas-solids flow is allowed to develop completely over a wider range of operating conditions. The fluidization gas used in the experiments was air, which was at ambient temperature and pressure, and supplied by a Roots-type blower. The paniculate materials used in this study were sand and fluid catalytic cracking (FCC) catalyst. The pressure gradient, local solids concentration and the local solids velocity were measured.According to the experimental data, the axial and radial distributions of the hydrodynamic characteristics (solids holdup, solids velocity, and solids circulating flux) were studied. The results showed that there are two types of core-annulus structures. The characteristics of the first type structure are that all of the solids in the bed bottom section flow upwards, and in the bed top section, the core solids flow upwards, but the annulus solids flow downwards. The characteristics of the second type structure are that all of the solids in the full bed are flow upwards.An artificial neural network model was set up to simulate the hydrodynamics of the gas-solid flow in the gas-solid circulating fluidized bed. The output of the model consists of five hydrodynamic parameters-cross section average solids holdup, local solids holdup, local solids circulating flux, cross section average solids velocity, and local solids velocity. The model can well simulate and predict full bed hydrodynamics of the gas-solid flow. Moreover the model can easily be used, which is significant in the industry application.Chaos theory was used to analyze the time series of solids holdup fluctuation,and consequently Kolmogorov entropy was calculated to describe the hydrodynamic characteristics of the gas-solids flow behavior in the riser. The results showed that Kolmogrov entropy could describe the two types of core-annulus flow structures mentioned above. According to the marked characteristics of Kolmogorov entropy at radial direction in the two types of flow structure, the flow regions were identified respectively.Finally, a prediction model based on combination of chaos with artificial neural network was proposed to predict the chaotic time series of solids holdup fluctuation. The results showed that the proposed model was an effective method in predicting the local solids holdup fluctuation and it was better than the traditional neural network prediction method. The predictability of the hydrodynamics of the CFB riser is possible only in a short time, which accords with that the gas-solid circulating fluidized bed is a chaotic system.