Bifurcation analysis of bubble dynamics in fluidized beds.

We use a low-dimensional, agent-based bubble model to study the changes in the global dynamics of fluidized beds in response to changes in the frequency of the rising bubbles. The computationally based bifurcation analysis shows that at low frequencies, the global dynamics is attracted towards a fixed point since the bubbles interact very little with one another. As the frequency of injection increases, however, the global dynamics undergoes a series of bifurcations to new behaviors that include highly periodic orbits, chaotic attractors, and intermittent behavior between periodic orbits and chaotic sets. Using methods from time-series analysis, we are able to approximate nonlinear models that allow for long-term predictions and the possibility of developing control algorithms.

Dynamic interacting bubble simulation (DIBS): an agent-based bubble model for reacting fluidized beds.

In this paper we explore the global dynamics of an agent-type model for bubbles in gas-fluidized beds and demonstrate that these features are consistent with experimentally observed behavior. The model accounts for the simultaneous interactions of thousands of individual bubbles and includes mass-transfer and first-order reactions between the gas and solids so that the impact of the dynamics is reflected in reactant conversion. We start with model parameters that have been demonstrated to produce time average behavior consistent with experimental reactor measurements. By observing the temporal variations of spatially averaged bubble properties, we are able to clearly distinguish the onset of global low-dimensional features that appear to be consistent with previous observations. The most prominent of these features is a large-scale oscillation that exhibits intermittency with power-law scaling in the vicinity of a critical gas flow. We show that the oscillation occurs as the result of a globally synchronized horizontal movement of the bubbles toward the center of the reactor. The oscillation appears to be consistent with the occurrence of the so-called "slugging" phenomenon, which is known to have large effects on fluidized bed reactor performance. Although this model can clearly be further improved, its success in replicating several of the key features of slugging indicates that this approach can serve as a useful tool for understanding and possibly controlling fluidized bed dynamics. We also conjecture that this model may be useful for more generally understanding the occurrence of global features in high-dimensional, multi-agent systems.