Full-Loop Study of a Pilot-Scale Dual Fluidized Bed Cold Flow System: Hydrodynamic Simulation

Computational particle fluid dynamics (CPFD) simulations of three-dimensional cold-flow gas-solid flow were performed for a pilot-scale dual-circulating fluidized bed (DCFB) using Barracuda Virtual Reactor® 17.4.0. This DCFB system is a fluidized bed gasifier used in a calcium-loop gasification process to produce a hydrogen-rich gas with CO2 capture. The effects of different bubble bed inlet airflow rates and different solid inventories on the full-loop pressure distribution and particle concentration distribution were compared. The increase in the bubble bed inlet airflow rate significantly increases the pressure in the bubble bed and lift pipe, and the increase in the solids inventory significantly increases the pressure where the solids are tightly packed. For long-term operation, the appropriate combination of parameters should be found to maintain a stable particle seal at the loopseals.


Introduction
The conversion of low-cost biomass materials into high-quality syngas has been attracting extensive attention.Dual fluidized bed technology has great potential for biomass gasification [1].Essentially, dual-circulating fluidized bed (DCFB) is an advanced technology in the field of fluidized beds.The complex structure of DCFB brings various difficulties in experimental studies.The design of DCFB should consider internal particles, flow fields, and various local dimensions to adapt to multiple operating conditions.Observation of particle dynamics requires advanced technologies such as capacitance tomography and laser doppler velocimetry [2].In addition, once the experimental bench is built, it is difficult to change its geometry, the position of each feed and discharge port, or other parameters.The damage caused by design errors is serious.Therefore, DCFB technology has increased the demands for advanced computational tools in design, evaluation, and scale-up, with its application in emerging fields such as biomass gasification.Computational fluid dynamics (CFD) is such a suitable, well-used and efficient tool [3].Modeling and simulation of fluidized beds by CFD is a solution to cope with the above-mentioned drawbacks.CFD simulation can obtain accurate data which is difficult to obtain by experimental observation.With the continuous improvement of computational performance and knowledge of the fluid-solid phases field, CFD will become more realistic and efficient.
Euler-Euler (EE) and Euler-Lagrange (EL) are two basic numerical calculation methods for gas-solid two-phase flow simulation.With the EE method, the gas-solid phases are considered to be continuous and fully interpenetrated.With the EL method, the trajectories of the particles are calculated separately.Although the empirical formulations related to fluid dynamics for EE simulations are relatively well established and some of their results are satisfactory, the discrete nature of the particles is not considered [4].The discrete particle method (DPM) under the EL approach has a computational cost of 80% of the total computational cost for the collision and motion of the particles.Hardware costs and time become expensive when applying the DPM method in medium to large fluidized beds.The multiphase particlein-cell (MP-PIC) method is an improved discrete phase algorithm that aggregates particles into a computational package and computes the entire particle size distribution, which greatly improves computational efficiency.
To accelerate the simulation calculation rate, numerical simulations are usually performed locally for CFB or DCFB in literature.It has been shown that the fluidized bed localization such as the interior of the lift tube is affected by the inlet and outlet particle flow parameters [5].This makes local simulations that require specific boundary conditions difficult to achieve accuracy and adapt to changes in inlet and outlet parameters.To improve the computational precision and to consider the interaction of all local structures, full-loop CFB simulations are gradually increasing.The MP-PIC model allows the simulation of particle flows in medium and large experimental benches due to its excellent computational efficiency.However, there is a lack of MP-PIC simulations for medium or large systems, which tend to be closer to the ideal industrial production process.Thapa et al. [6] modelled MP-PIC simulations of lab-scale CFB and analyzed the choice of the location of each inlet on the full bed.
The modeling of cold-flow full-loop DCFB systems using CPFD simulations is not abundant enough.In order to enrich this aspect of research, this study applies the MP-PIC model to perform full-loop simulations of cold-flow, pilot-scale DCFB systems with the aim of analyzing the effects of solid packing volume and inlet air volume on their fluid dynamics.The hydrodynamic analysis of cold flow can provide data support for subsequent thermal state experiments.

Experimental
The reactor structure of the pilot-scale experiment is a DCFB. Figure 1 demonstrates the overall structure of the reactor.The reactor consists of a gasifier, a regeneration reactor, a riser, two cyclones, and three loopseals.The diameter of the gasifier bubble bed is 0.75 m and the height is 6 m.The diameter of the lift pipe is 0.19 m and the height is 11.5 m.The diameter of the regeneration reactor is 0.29 m and the height is 19.3 m.Dolomite was chosen as the circulating material, and grain size was sieved from 0.15 to 0.3 mm.It will be used as the calcium-based carbon adsorbent in the subsequent hot-state experiments.The particle size range after sieving of dolomite particles falls into the B particle class of the Geldart classification.The true density and bulk density of the dolomite used were 2750 kg/m 3 and 1600 kg/m 3 , respectively.During system operation, an appropriate amount of supplemental dolomite is fed into the bubbling bed through the inlet to achieve a stable bed material content.The gas enters the circulation line tangentially, thus serving as the internal cyclone at the two gas outlets.The secondary cyclone in the actual experiment is omitted in the simulation with the outlet bed material circulating to the inlet to improve the calculation speed.

CFD model setup
Numerical CPFD simulations of three-dimensional cold-flow gas-solid flow were performed for a pilotscale DCFB using Barracuda Virtual Reactor® 17.4.0.The particles were tracked for a solution using the MP-PIC model in a Lagrangian framework, which grouped the dispersed particles together.MP-PIC method is novel for particle calculations, making the calculations efficient and accurate.In the current study, the research focuses on the study of fluid dynamics in fluidized beds.Chemical reactions such as biomass particle gasification are neglected in the simulations.It has been well documented that the mass of biomass particles is usually less than 5% of the total particle mass.Therefore, the particle flow of a DCFB is mainly influenced by the filling bed material.The hydrodynamic behavior of the biomass particles in the full cycle is negligible.In order to prevent bed material particles from running out of the cyclone outlet, the two cyclone outlets and the bubble bed inlet are linked by a BC connector to achieve an internal circulation of bed material.
The grid diagram is displayed in Figure 1(b).The total number of grids is 2652756.The mesh size is about 100 times the size of the particles, and the grid encryption is performed locally for the cyclone and the ring seal, which is acceptable for the MP-PIC calculation model, due to the unique particle aggregation algorithm of MP-PIC.Three kinds of grid divisions were established with total grid numbers of 1210561, 2652756, and 4003856, respectively.When the grid number is 1210561, the local mesh quality of the cyclone separator and the annular seal is poor and the model is distorted.There is no significant difference in pressure distribution and particle distribution after reaching stability by running for 10 s under the working conditions with grid numbers of 2652756 and 4003856, respectively.Therefore, the grid number of 2652756, which is faster in calculation speed, is chosen.

Governing equations
The strategy solves the continuity equations of the fluid with a modified Eulerian grid and solves the Lagrangian particles using sub-grids.The continuity equation of a compressible fluid can be expressed as [7]: where θ g , ρ g , u g , p, τ g and F are the density, volume fraction, gas velocity, pressure, Reynolds stress tensor and momentum exchange rate, respectively.The particle phase of the MP-PIC scheme is controlled using the Liouville function f, which can be expressed as [7]: where  is the divergence operator of the particle velocity component, and u s is the solid phase velocity.The computational equation of particle motion is expressed as [7]: where D s is the interphase drag coefficient.The drag model is the key parameter that controls the simulated particle dynamics.The resistance of the particles in the dense and dilute phase gas streams is different.Empirical formulas on the resistance model were obtained in the experimental study and were able to explain this phenomenon.The drag models proposed by Wen, Yu and Ergun have been widely used and proved to be accurate.The Wen-Yu/Ergun blended equation is applied to MP-PIC simulations as a benchmark model: ( ) where θ s, CP is the close-pack particle volume concentration.

Results and discussions
Figure 2 shows the gauge pressure distribution as a function of measurement point height in the DCFB system after the stable operation.The locations of all measurement points are consistent with those constructed for the later experiments.For all four operating conditions given here, the maximum pressure fluctuation is in the lower part of the bubbling bed, while there are local high pressures at three loop seals.In contrast, the pressure and pressure drop are smaller near the cyclone to maintain its solids Comparing the results of B3F25, B6F25 and B9F25, it is shown that the pressure drop of the gasification reactor decreases (P1-P3) and the pressure of the riser tube increases by increasing the amount of gas introduced to the bubble bed, which indicates that the retention rate of particles in the riser tube decreases.It is observed that in the lower loop seal, the bubble bed gas volume has almost no effect on its pressure drop, which is caused by the large number of particles retained between the bubble bed and the regenerative reactor and the stable particle circulation rate.In contrast, in the middle loopseal (P5-P2) and upper loopseal (P10-P2), the pressure drop is more influenced by the bubble bed gas volume, so the fluidizing air needs to be properly adjusted to stabilize the direction and rate of solids circulation in the case of a stable particle seal.Comparing the results of B6F25 and B6F25S2, it is shown that the solid inventory has a significant effect on the high-pressure region in the bubbling bed.The increase in solid inventory directly increased the height of the particles in the bubbling bed, while in the B6F25S2 with a pre-filled particle volume of 2000 kg, the high-pressure region significantly exceeded the return duct opening of the middle loop seal.This also affected the particle circulation of the middle loop seal, and in the comparison of the two operating conditions it was found that the increase in the pre-filled particle volume reduced the mass flow rate of the middle loop piping.Therefore, for operating conditions with different solid inventory amounts, fine tuning the fluidizing air of the sealing loop can maintain a stable solids cycle.Proper adjustment of the loop inlet height position in the bubbling bed is also a way to avoid the effect of particle filling height on the solid circulation.Starting from 10 s, the flow rate of each channel stabilized, so the condition of the particle volume fraction from 10 s to 20 s was recorded.Figure 3 shows the instantaneous particle distribution at t=10 s, 15 s, and 20 s for bubble bed inlet rates of 0.3 m/s, 0.6 m/s and 0.9 m/s, respectively.The concentration of solids is relatively small in the upper part of the bubbled bed, because the local apparent gas velocity is not sufficient to entrain the large amounts of solids.Comparing the degree and area of particle aggregation within the bubble bed, the elevated gas flow rate significantly increased the degree of fluidization within the bubble bed.Due to the relatively tightly sealed channels of the loop, the bubble bed inlet gas rate had a weak effect on the overall solids volume distribution.At the inlet flow rate of 0.3 m/s, the flow rate of each channel was stable while there was a significant and continuous decrease in the particle seal height in the upper loopseal and middle loopseal during 10 s-20 s.In contrast, there is no obvious behavior of continuous decrease of seal height when the inlet air is 0.6 m/s and 0.9 m/s.This is mainly because the inlet air rate affects the pressure drop between the bubble bed and the loopseal.At the same time, the particles did not adequately replenish the particle inventory at the loopseal due to the continuous particle outflow from the cyclone.In long-term operation, the inlet rate at the bottom of the bubble bed affects the pressure, and the pressure variation affects the solids circulation flow rate, gradually affecting the height of the particle seal in each loop seal.Therefore, for each inlet rate, the settings of fluidized air and cyclone separator at the loop seal should be adjusted appropriately to ensure that the loopseal maintains a certain degree of tightness, thus ensuring long-term stable operation of the system.Figure 3. Instantaneous particle distribution at t=10 s, 15 s, and 20 s for bubble bed inlet rates of 0.3 m/s, 0.6 m/s and 0.9 m/s, respectively.

Conclusions
The cold gas-solid flow in a pilot-scale DCFB system was investigated by CPFD simulations.A 3D fullloop model based on the experimental setup was developed to investigate the key parameters affecting the fluid dynamics.The above analyses are summarized as follows: (1) The increase in the inlet rate at the bottom of the bubble bed raises the pressure of the bubble bed and the lift pipe, which has a direct effect on the circulation flux of the loop directly linking the bubble bed and the lift tube.

NESP-2023
Journal of Physics: Conference Series 2592 (2023) 012033 (2) The increase in solid inventory significantly increases the high-pressure region of the bubble bed and reduces the circulation flux of the loop connected to the high-pressure region.
(3) In long-term operation, the effect of pressure changes on the solid circulation flux caused by the air inlet rate at the bottom of the bubble bed gradually affects the height of the particle seal in each loop seal.
In general, the solid inventory and the gas flow rate of each inlet should be properly adjusted to keep the particle circulation stable.This study will provide some reference for the subsequent experimental studies and keep optimizing the accuracy of various parameters in the full-loop simulation and searching for favorable parameter combinations, so that the predictability of the model can actually become a useful tool.

Figure 1 .
Figure 1.a) Sketch of cold-flow DCFB system.b) Computational grid c) Full-loop with flux tracking planesThe parameter settings and operating parameters of the full-loop cold flow simulation are shown in Table1.The drag model was set with Wen-Yu/Ergun drag coefficient, based on[7].

Close
separation function.Differential pressure and gas feed are the reasons for the smooth and stable solid circulation of the DCFB system.

Figure 2 .
Figure 2. Average gauge pressure distribution for the simulations under four operating conditions (average pressure during 19 s-20 s).Instantaneous pressure distribution at t=20 s for each case with measurement points labeled.

Table 1 .
Simulation setting and operational parameters.