Impact of Fluidized Bed Hydrodynamics on the Distribution of Liquid Sprayed into the Bed

The Fluid Coking Technology^TM is one of the most advanced upgrading processes for heavy oil reserves. In Canada, for example, it is used to upgrade around 15-20 % of heavy oil production to a light, synthetic crude. In a Fluid Coker, preheated bitumen (350 ℃) is sprayed through steam-gas atomization spray nozzles into a fluidized bed of hot coke particles (500-550 ℃) to be thermally cracked to condensable vapors, non-condensable gases and solid coke. However, when spraying the bitumen into a bed of hot coke particles, a portion of the liquid is trapped in wet agglomerates formed with the solid coke particles. Coke particles flowing down from the spray region go through a stripper section before being conveyed to a burner vessel, where they are reheated. Wet agglomerates entrained with the particles flowing through the stripper are responsible for fouling of the stripper sheds, which causes major operating problems. It is, therefore, essential to minimize the formation of wet agglomerates.

This research focuses on reducing agglomerate formation and enhancing agglomerate breakup by modifying the bed hydrodynamics in the spray region. It investigates the impact of changes in the distribution of gas bubbles that can be achieved by either modifying the distribution of the fluidizing gas or adding internals such as baffles.

This study uses a safe system that has been found to simulate agglomerate formation and breakup in a Fluid Coking process: an aqueous solution of Gum Arabic is sprayed into a fluidized bed of sand particles maintained at 120 °C. Experiments were conducted in a column with a rectangular cross-section (1 m × 0.5 m × 2.28 m). Liquid was atomized into the bed using nitrogen and a scaled-down version of a typical commercial spray nozzle. Liquid was injected for about 11 s, the bed was dried and its particles were screened to recover the agglomerates, which were then sieved to get their size distribution. The recovered agglomerates were analyzed to determine their Gum Arabic concentration, from which the proportion of the injected liquid that was trapped in agglomerates could be determined. Experiments were conducted with superficial velocities ranging from 0.18 m/s to 1.0 m/s.

The flow patterns of the gas bubbles were determined with an array of triboprobes. Triboprobes consist of metal rods connected to the ground. When sand particles carried by the bubble wakes hit a probe, they strip electrons from its surface, inducing an electrical current that can be monitored. The main advantage of triboprobes, when compared to optical fibers or capacitance probes, is that they can be used at high temperature and with high gas velocities.

The gas distributor was equipped with 20 gas nozzles in two parallel rows, whose flow could be controlled individually to modify the gas bubble distribution. Both amplifier and resistance have been applied and compared for triboelectric method. By applying the resistance, the voltage ( is recorded as the raw signal. And the Data acquisition is protected with transient-voltage-suppression (TVS) diodes from the damage of high-voltage from the fluidized bed. With this improvement, we greatly lower the cost of recording data from multi local positions contemporaneously.

By changing the gas distributor configuration, opening 10 nozzles each time, three extreme cases of gas distributions are obtained. By opening 10 nozzles near the back-bed wall, an even gas distribution is formed, which is flatter than the gas distribution from a conventional fluidized bed due to bubble coalescence. By opening 10 nozzles close to the western side bed wall, the bubbles would concentrate on the western bed side, which is called western case. Similarly, we can have the eastern case when the bubbles concentrate on eastern side bed wall. For these three extreme cases, the results have been verified with both amplifier and the resistance method.

The gas bubble distribution was also modified by inserting baffles in the fluidized bed. An open-ended right triangle shape with an internal angle of 45 ° having the same width as the bed (0.18 m x 0.18 m x 0.1 m) is set as the basic baffle geometry. To simulate the frusto-conical baffle having a central aperture applied in industry, two lengths of O-shaped flux-tubes are put at the center of the basic geometry baffle separately; one is the same length as the bottom edge of the baffle, the other one is half length of the full-length flux-tube.

The results from triboelectric method show that all three geometries of baffle would redirect the bubbles to the baffle side. However, there’s no great impact on the bubble profiles from different baffle geometries, where the measuring level is right above the spray. When applying a basic geometry baffle, by changing the gas distributor configuration, for all the three cases, even, western, and eastern case, the bubbles are redirected to the baffle side, while keeping the even case is in-between of the western and eastern case at spray region.

The results from Gum Arabic method shows that by changing the gas distributor, the amount of water trapped in agglomerates can be reduced, with the best results obtained when the gas is evenly distributed at the spray level. This could be because this configuration both provides a high solids entrainment into the spray jet cavity, near the nozzle tip, and high rate of agglomerates break-up near the spray end.

The basic geometry baffle decreases the portion of trapped liquid for all gas velocities, gas distributor configurations, and vertical distance from the spray nozzle. However, the geometry of the flux-tube baffle needs to be optimized.