(465e) Use of Multiphase CFD Simulations to Address Erosion Problems in a FCC Regenerator at the LyondellBasell Houston Refinery

In March 2008 LyondellBasell's Houston Refinery FCC unit experienced an unplanned shutdown due to a control failure that brought down the wet gas compressors. Though the regenerator was not responsible for the unplanned outage, it was inspected as a matter of routine. This inspection showed serious erosion of some regenerator internals, particularly horizontal support structures and the exterior of the cyclone diplegs (Figure 1). It was observed that the erosion was concentrated in only one particular quadrant of the regenerator. The unit was repaired and quickly placed back in operation. It was realized that considerable investigative work lay ahead to ascertain the root-cause of the erosion and develop design modifications to ensure reliable future operations. A scheduled shutdown was then planned for early 2011 to install the appropriate modifications to address the erosion issues.

This regenerator is an enormous vessel approximately 100 feet tall and over 50 feet in diameter filled with complex internals, with air delivered by a steam-turbine driven main air blower. The entire FCC unit has a catalyst inventory of approximately 1400 tons, of which the majority resides within this regenerator. The plan was to model the fluid-particle flows before and after the recent hardware revamp, explain what had caused the erosion damage, and assess potential redesigns to ensure reliable operations. LyondellBasell decided to employ the commercial Barracuda® CFD technology to understand more fully the fluid-particle flows within the regenerator. Traditional commercial computational fluid dynamics (CFD) packages are designed for single-phase flow, either gas-only or liquid-only flows. Extensions over the 40 years or so that CFD has been popularized have allowed these packages to handle some narrowly defined particle-fluid problems. However, modeling highly coupled solid-gas problems at an industrial scale required a new technology. Built from the ground up for solid-fluid problems, the Barracuda simulation package provided the specific technical capabilities required to undertake this regenerator simulation effort with confidence. The numerical method behind Barracuda is termed CPFD®, or Computational Particle Fluid Dynamics. CPFD is a simulation approach combining a Lagrangian method for the particles with an Eulerian method for the fluid (gas or liquid) flow. In addition to helping users understand the particle & fluid flow within their applications, Barracuda offers full thermodynamics and chemistry capabilities and is well suited to modeling large industrial particle-fluid systems at full scale, with or without chemical reactions. Barracuda is used for a wide variety of particle-fluid flows but is most commonly used in fluidized beds such as this regenerator.

The Barracuda regenerator model was done at full-scale and in three-dimensions. Geometry and boundary condition locations are shown in Figure 2. The simulations showed immediately a significant imbalance of the ?spent-cat' particles flowing in the return line; the imbalance was a result of the 90-degree J-bend in the return line. A majority of the particles are ?thrown' to the outside of the bend (the south side of the regenerator) as the pipe turns upwards into the regenerator; this asymmetrical flow forces the air to favor the north side by an enormous amount, a 10:1 margin. The result is a high-velocity, particle-laden jet of gas that impinges on the internals, primarily on the north side of the regenerator. Figure 3 illustrates the difference in time-averaged particle volume fraction between the distributor installed in 2007 (as-built) versus an alternative design for the same air flow. The alternative design showed a much more even distribution that tended to stay toward the center of the vessel and away from the cyclone diplegs that ring the interior of the vessel.

Comparisons were made for the different distributor designs at the same air flow rate: the as-built case, and two different footprints of the alternative design. One of the alternative designs had the distributor placed higher in the vessel to allow for improved flow profile development up the spent-catalyst riser, but without increasing backpressure on the blower. This will allow the regenerator to use higher air flow rate capability of the blower (if necessary) that are essential to maintaining the improvements in conversion for the FCC unit as a whole. The new design is also intended to ameliorate the impact of the increased air flow and the 90-degree bend in the spent cat return line. The comparison of particle volume fraction for the entire vessel, Figure 4, clearly shows the asymmetry of the flow for the as-built case and shows more uniform distribution for both of the alternative design distributors. Figure 5 shows the particle residence time for particles entering the vessel. The as-built distributor allows a greater amount of newly unregenerated spent catalyst entering the vessel to be thrown into the freeboard, which would be detrimental to controlling afterburn. The alternative designs both do a better job of keeping newly entering spent catalyst in the bed. While the volume fraction and residence time plots show the two alternative designs to be similar, the isovolume of time-averaged velocity, shown in Figure 6, shows greater differences. There is some asymmetry for all the distributors but the alternative design (with higher placement of the distributor) keeps the inlet plume in the center of the vessel and away from the cyclone diplegs where the erosion was most severe. Wear is highly dependent on particle velocity and the isovolume of wear intensity in Figure 7 shows this alternative design is effective at keeping the wear centered in the vessel and away from the outer two rings of cyclone diplegs. The centermost diplegs, where the wear is heaviest, may be protected with refractory during the next planned outage.

CPFD simulations with Barracuda showed that simply returning to the pre-2007 distributor would not solve the wear issues. Wear might be reduced below 2007 levels, but not sufficiently to provide a nominal 5-year period between overhauls. The understanding provided by CPFD simulations has allowed a more thorough evaluation of the alternative design with a very high likelihood of minimizing wear substantially. The new design not only slows the inlet flow, reducing the velocity of particle-laden jets impinging on internals, but also provides a much better distribution of the solids within the bed, greatly reducing the bias of gas flow to the north side of the unit, which was the cause of much of the erosion. Prediction of the most likely high-wear areas also allows for refractory to be placed on the exterior of cyclone diplegs where erosion is most likely, which with the proposed design would be the center cyclones only.

In summary, the initial Barracuda simulations of the as-built condition predicted wear patterns consistent with those found in the regenerator during the 2008 inspection. Simulations comparing the pre-2007 configuration to the as-built unit provided insight into the causes for this erosion. CPFD simulations of proposed new distributor designs allowed the most effective design to be selected in order to improve the operational life of the regenerator substantially without reducing the significant gains in conversion achieved by the recent FCC unit revamp. These advanced engineering simulations have allowed the refinery and the engineering contractor to move forward with the necessary regenerator modifications with reduced risk and greater confidence.