Sie sind auf Seite 1von 4

Communications

Symbols used
c d32 m m NTU x [mg/Nm3] [m] [kg] [kg/h] [] [] concentration Sauter-Diameter mass mass flow number of transfer units mass fraction

Numerical Simulation of Particle Wall Adhesion in Gas-solid Flows


By Eckart Heinl* and Matthias Bohnet

1 Introduction
Numerical research was implemented, concerning the interaction between a disperse and continuous phase in dilute pneumatic conveying. Prerequisites were established by Tsuji [1], Sommerfeld [2], Frank/Petrak [3], and Werninger [4]. They developed models for particle-wall collision and/or particle-particle collision and implemented them into their Computational Fluid Dynamic (CFD) program codes. Thereby, it became possible to describe particle behavior in a flow, particularly with regard to pressure drop and concentration distribution across the cross section. Common to this numerical research was the use of free flowing particles. Until now no numerical investigations existed concerning pneumatic conveying of adhesive particles. Particle wall adhesion involves an additional pressure drop that is neither wanted nor calculable. Furthermore, re-entrainment of adhering particles leads to variations in concentration and pressure along the conveying line. Both complicate the design of conveying lines. CFD can give a realistic impression of conveying using a suitable model of particle wall adhesion. This work introduces one possibility for modeling particle wall adhesion and its implementation in FLUENTs CFDcode.

Greek symbols d P j [C] [] [%] temperature conversion relative humidity

Indices aus Binder Ca(OH)2 ein F G Keime SO2 Susp U WS out binder Calcium hydroxide inlet liquid gas nuclei Sulfur dioxide suspension environment fluidized bed

References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] M. Trojosky, L. Mrl, Chem. Tech. 1991, 4, 141. S. Heinrich, L. Mrl, Chem. Eng. Proc. 1999, 38, 635. M. J. Hounslow, R. L. Rayll, V. R. Marshall, AIChE J. 1988, 34, 1821. M. J. Hounslow, AIChE J. 1990, 36, 106. A. A. Adetayo, J. D. Lister, S. E. Pratsinis, B. J. Ennis, Powder Technol. 1995, 82, 37. M. Nicmanis, M. J. Hounslow, AIChE J. 1998, 44, 2258. W. D. Goroschko, R. B. Rosenbaum, O. M. Todes, Neft i Gaz. 1958, 1, 125. H. Martin, VDI-Wrmeatlas, 7th ed., VDI-Verlag, Dsseldorf 1994. H. Groenewold, E. Tsotsas, in Proc. 11th Int. Drying Symposium, Halkichiki, Greece 1998, Vol. A, 192. D. Shi, VDI Fortschritt-Berichte, Reihe 3, Nr. 466, VDI-Verlag, Dsseldorf 1997. W. Kast, H. Klan, VDI-Wrmeatlas, 7th ed., VDI-Verlag, Dsseldorf 1994. L. Mrl, Ph.D. Thesis B, TH Magdeburg 1981. J. Rangelova, J. Dalichau, S. Heinrich, L. Mrl, Chem. Ing. Tech. 2001, 73, 1124. M. Peglow, M. Ihlow, L. Mrl, in Proc. Energy Forum, Varna, Bulgaria 2002, Vol. 1, 281.

2 Models
For the simulation, the commercial software code FLUENT 6 was used. Models describing particle wall adhesion, additional forces, and wall roughness were implemented via user defined functions (UDF). The horizontal pneumatic conveying of quartz powder (dP,50 = 3 lm, MP = 11.4 g/s) was calculated with the Euler-Lagrange approach, i.e., representative particle tracks were determined by integrating equations of motion resulting from the force balance of inertia, drag, gravity, and buoyancy1). Additionally, Saffmann and Magnus lift forces are considered. Also, there was the option of taking the electrostatic charge of the particles into account, via the Coulomb force. The particle wall adhesion was described with the van-derWaals energy that is necessary to let particles leave the wall
[*] Dr.-Ing. E. Heinl, Volkswagen AG, Letter box 011/1778, Group Research, Fuels and Lubricants, D-38436 Wolfsburg, Germany; Prof. Dr.-Ing. M. Bohnet, Institute of Chemical and Thermal Process Engineering, TU Braunschweig, Langer Kamp 7, D-38106 Braunschweig, Germany. List of symbols at the end of the paper.

This paper was also published in German in Chem. Ing. Tech. 2003, 75 (12), 1852.

______________________

1) Chem. Eng. Technol. 2004, 27, No. 11 DOI: 10.1002/ceat.200407037

 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1143

Communications after impact. Therefore, the energy balance at the particle wall collision was considered. It is implied that the kinetic energy of a particle diminishes if its energy is not sufficient to let the particle leave the wall after impact. This model was developed by Lffler/Muhr [5] and was applied here in order to additionally consider the electrostatic energy. This yielded a critical particle impact velocity as follows: wP;krit

3 Results
In Fig. 1, the mass of particles at the wall is depicted over time for a flow with Re = 82730 and a loading (l = MP /MF) of 0.2.

s    
2

(1)

$ h 2 2 edP 4p z
0

3 4H r

3 1 d3 p 2 2 e r e
P

2q 2

0 P

2z0 dP

P ;2

q2

P ;1

A particle is regarded as adhering in the current grid cell if the condition for adhesion is fulfilled. wP;1  wP;krit (2)

Figure 1. Impact of wall roughness and electrostatic charge of particles on particle wall adhesion.

Furthermore, it was ensured that particles can only enter grid cells that are not full of particles. Should the latter be the case, wall collision, and condition of adhesion were checked in the current grid cell. In order to shape CFD calculations, independent of the number of representative particles, only a certain amount of the particles in such a representative particle package was considered to adhere. In fact, only sufficient particles adhere at the wall to cover the current grid cell with one mono layer. An additional model enables the consideration of wall roughness. Therefore, the model developed by Werninger [4] and Triesch [6], which originates from Frank/Petrak [3], was adjusted to the particle size considered.

2.1 Grid and Boundary Condition of the Numerical Simulation A horizontal pipe flow (Di = 50 mm) was simulated. The Reynolds-Averaged-Navier-Stokes (RANS) equations were solved with a realizable k-e model. In order to solve the equations in the boundary layer, a two layer Wolfshtein model was used [7]. The latter involves very high grid resolution in the boundary layer and, therefore, a high demand for CPU time. Hence, only a 15 cross-section of the pipe was taken into account. Tangential boundary areas were defined as symmetric, i.e., particles keep on moving in the calculation volume retaining there velocity. Allocation of mass to the representative particle packages was performed with a Rosin-Rammler particle distribution [8] using measured particle size values (dP,min = 0.5 lm, dP,max = 11 lm, dP,50 = 3 lm). In order for the roughness model to be valid for a smooth steel pipe of the wall, the values of Frank/Petrak [3] are used: (hR = 5 lm, lR = 50 lm). 1144

After an initial steep increase, the mass of adhering particles for all cases shown only exhibits a small further increase. This behavior is causes by the model of layer growth. The thicker the layer, the bigger the velocity at which particles collide with the layer. As a result, the probability of matching the condition for adhesion decreases (Eq. (2)). Neglecting the electrostatic charge of the particles, the seemingly bigger adhesion in the first 2.5 s, is derived from the coarse grid resolution; particles enter the same cell wall more frequently. Considering the electrostatic charge of the particles results in a bigger critical particle velocity and hence in faster adhesion. At the location of collision, particles stick faster and a layer is formed. Successive particles, which would collide with the same cell wall, are prevented from doing so by the layer model. They collide with the layer and not with the wall. At the layer, the probability of adhesion is smaller due to the higher particle velocity. If electrostatic charge is neglected, adhesion occurs only after a huge number of wall collisions, which decelerate the particle. In this way, particles are able to distribute evenly over the pipe wall. This behavior mainly results in a bigger particle wall adhesion, if electrostatic charge is neglected. After approximately two and a half seconds the real behavior is described: the increased particle wall adhesion for the case where the electrostatic charge of the particles is considered. In Tab. 1 the mean values for the local loading as well as the total and radial particle velocities in the boundary layer are shown for a smooth and a rough wall. Obviously wall roughness slows down particles more in the boundary layer, yielding a bigger adhesion. Additionally, more particles are accelerated towards the middle of the pipe (negative sign), which decreases the mean of the radial particle velocity. In contrast, the mean radial particle velocity is oriented towards the pipe wall (for a smooth wall).
http://www.cet-journal.de Chem. Eng. Technol. 2004, 27, No. 11

 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
Table 1. Loading, total, and radial particle velocity in the boundary layer. loading [] smooth wall rough wall 0.65 0.48 total particle velocity [m/s] 15.11 13.5 Radial particle velocity [m/s] 0.13 0.03

The pressure drop can be described by a single flow in a rough pipe, where the resistance coefficient, k, is described by the approach of Colebrook:   1 2ks 18p :7 p 1; 74 2log (4) Di Re k k with pmeasure 2 Di u2 rL

Hence, more particles leave the boundary layer for a rough wall, whereby particles accumulate in the boundary layer for a smooth wall. The results, regarding particle wall adhesion for different air velocities, are depicted in Fig. 2.

(5)

Figure 2. Impact of air velocity (loading) on particle wall adhesion.

The loading decreases with an increased air velocity and a constant particle mass flow rate. In addition, a higher air velocity involves a higher particle velocity. Both imply a decrease in particle wall adhesion. The first is due to the lower particle concentration in the boundary layer, the latter is due to a lower probability of adhesion, which results from the higher particle velocity and the constant critical particle velocity, Eq. (1). This value can only be estimated since no suitable inline particle wall adhesion measurement device was available. The simulated flow pressure drop was measured in a test rig at a constant mass flow rate (quartz powder, MP = 11.4 g/s, dP,50 = 3 lm). The inner pipe diameter is 50 mm. Measurements covered 810 min. The pressure drop increases after conveying begins, although the feeding of particles and the air velocity were held constant. Furthermore, the pressure drop stays on a higher level than before conveying after stopping the particle feed. This behavior is caused by the build up of a particle layer at the pipe wall, which involves a decrease in the cross section and an alteration in wall roughness. In order to estimate particle wall adhesion the following two assumptions were made: 1. the layer thickness is equalized with the roughness height ks, 2. the increase in the pressure drop is solely due to the increase in ks; the part which is caused by the conveying of particles is neglected.
Chem. Eng. Technol. 2004, 27, No. 11 http://www.cet-journal.de

the roughness height, ks, can be determined. As a result, the height of the particle layer, which is obtained after conveying, can be estimated. Considering the time, this results in a mean particle layer growth of 0.7, 0.08, and 0.01 lm/s for air velocities of 20, 25, and 30 m/s, respectively. In the simulation, the total mass of the adhering particles is known. This is transformed into a layer thickness using the area of the pipe wall and the bulk density of the quartz powder. Together with the time of conveying, this yields a mean growth of layer per unit time. For the simulation, it results in a mean layer growth of 2.28, 2.17, and 1.89 lm/s with air velocities of 20, 25, and 30 m/s, respectively. Due to the short simulated flow time the steep increase in the first second has more severe consequences than in the experimental data. This implies bigger simulation results than the experimental estimations. However, the decrease of adhesion that occurs with a decreasing loading and increasing air velocities is determined correctly. Experimental data, after 10 s of conveying, are used to compare between simulated and measured pressure drops. Qualitatively, the pressure drop is simulated correctly, as can be seen from Fig. 3.

Figure 3. Comparison between simulated and measured pressure drop.

As Sommerfeld determined, the frequency of wall collisions is lower for a rough wall than for a smooth wall, Fig. 4, [9]. However, this does not automatically result in a smaller pressure drop, as can be seen from Fig. 3. The pressure drop in the case of a rough wall goes beyond that of a smooth wall, although the difference is rather marginal. Additionally, Fig. 3 makes it clear that the pressure drop calculated by the simulation is too large regardless of the wall model
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1145

Communications ks L l k MP MF MP l pmeasure qP,i Re r rP u wP,krit wP,1 z0 [m] [m] [m] [] [kg] [kg/s] [kg/s] [] [Pa] [C] [] [kg/m3] [kg/m3] [m/s] [m/s] [m/s] [m] height of roughness length of pipe distance of current particle from the wall coefficient of restitution for the pipe wall particle mass air mass flow rate particle mass flow rate loading measured pressure drop charge of particles, i = 1, 2 before and after wall collision Reynolds number air density particle density air velocity critical particle velocity particle velocity before wall collision wall distance at wall contact

Figure 4. Number of wall collisions at a smooth and a rough wall.

(smooth or rough). Even for the single air flow, a difference exists for velocities of 25 and 30 m/s. This is probably due to a rather poor tangential and axial grid resolution compared to the radial resolution. These constraints in grid resolution were chosen in order to reduce the need for CPU time.

References
[1] [2] [3] [4] [5] [6] [7] [8] [9] Y. Tsuji, Y. Morikawa, T. Tanaka, N. Noakatsukasa, M. Nakatani, Int. J. Multiphase Flow 1987, 13, 671. M. Sommerfeld, KONA 1998, 16, 194. T. Frank, D. Petrak, in Proc. IV International Conference on Pneumatic Conveying, Budapest, Hungary 1990. C. Y. Werninger, Dissertation, TU Braunschweig 1996. F. Lffler, W. Muhr, Chem. Ing. Tech. 1972, 44 (8), 510. O. Triesch, M. Bohnet, Powder Technology 2001, 115, 101. M.Wolfshtein, Int. J. Heat Mass Transfer 1969, 12, 301. FLUENT6: Users Guide3, Ch. 14, 2001. M. Sommerfeld, Int. J. of Multiphase Flow 2003, 29 (4), 675.

4 Conclusion
The implementation of a model is introduced, describing particle adhesion at a pipe wall in the pneumatic conveying of very fine particles, using existing software code (FLUENT 6). Basically, it could be shown that this model is suitable, in combination with the Euler-Lagrange approach, for describing the continuous and dispersed phase for the simulation of particle wall adhesion in a pipe. The influence of air velocity and loading on particle wall adhesion is determined qualitatively, in an accurate manner. Qualitatively, the modeling of wall roughness results in the correct values for pressure drop and particle wall collision frequency. The quantitative statement of the simulation has to be scrutinized since the grid resolution is obviously far from optimized.

This paper was also published in German in Chem. Ing. Tech. 2003, 75 (9), 1286.

______________________
Rational Catalyst Design of Methanol Synthesis Catalysts*
By Melanie Kurtz, Natalia Bauer, Hagen Wilmer, Olaf Hinrichsen**, and Martin Muhler

Acknowledgements
The authors would like to thank the DFG for supporting this project financially in the scope of the SPP 1062.
Received: August 17, 2004 [CET 7037]

1 Introduction Symbols used


Di dP dP,50 e e0 Fr H h$  1146 [m] [m] [m] [] [As/Vm] [] [] [Nm] inner pipe diameter particle diameter mean particle diameter coefficient of restitution for collision relative permittivity Froude number hardness of wall material Lifshitz-van-der-Waals constant Since the 1960s, the Cu/ZnO/Al2O3 catalyst prepared by the co-precipitation method has been employed in ICI pro [*] This article has been presented at the Annual Meeting of GVC and DECHEMA, September 1618, 2003, Mannheim, Germany. [**] Dr. M. Kurtz, Dipl.-Chem. N. Bauer, Dr. rer. nat. H. Wilmer, Priv.-Doz. Dr.-Ing. O. Hinrichsen (author to whom correspondence should be addressed, O.Hinrichsen@techem.ruhr-uni-bochum.de), Prof. Dr. rer. nat. M. Muhler, Laboratory of Industrial Chemistry, Ruhr-University Bochum, D-44780 Bochum, Germany. DOI: 10.1002/ceat.200407032 Chem. Eng. Technol. 2004, 27, No. 11

 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Das könnte Ihnen auch gefallen