Centrifuge Separation Gas Liquid

FIG. 14-108 (a) Liquid entrainment from the bottom of a vessel by centrifugal flow. (Rietema and Verver, Cyclones in Industry, Elsevier, Amsterdam, 1961. ) (b) Gas-outlet skirt for liquid cyclones. (Stern et al., Cyclone Dust Collectors, American Petroleum Institute, New York, 1955. )

Reentrainment is generally reduced by lower inlet gas velocities. Calvert (R-12) reviewed the literature on predicting the onset of entrainment and found that of Chien and Ibele (ASME Pap. 62-WA170) to be the most reliable. Calvert applies their correlation to a liquid Reynolds number on the wall of the cyclone, NReL = 4QL /hivL, where QL is the volumetric liquid flow rate, cm3/s; hi is the cyclone inlet height, cm; and vL is the kinematic liquid viscosity, cm2/s. He finds that the onset of entrainment occurs at a cyclone inlet gas velocity Vci, m/s, in accordance with the relationship in Vci = 6.516 — 0.2865 ln Nrc,l.

Reentrainment from the bottom of the cyclone can be prevented in several ways. If a typical long-cone dry cyclone is used and liquid is kept continually drained, vortex entrainment is unlikely. However, a vortex breaker baffle in the outlet is desirable, and perhaps a flat disk on top extending to within 2 to 5 cm (0.8 to 2 in) of the walls may be beneficial. Often liquid cyclones are built without cones and have dished bottoms. The modifications described earlier are definitely needed in such situations. Stern, Caplan, and Bush (Cyclone Dust Collectors, American Petroleum Institute, New York, 1955) and Rietema and Verver (in Tengbergen, Cyclones in Industry, Elsevier, Amsterdam, 1961, chap. 7) have discussed liquid-collecting cyclones.

As with dust cyclones, no reliable pressure-drop equations exist (see Sec. 17), although many have been published. A part of the problem is that there is no standard cyclone geometry. Calvert (R-12) experimentally obtained AP = 0.000513 pg(Qg/hiWi)2(2.8hiwi/do), where AP is in cm of water; pg is the gas density, g/cm3; Qg is the gas volumetric flow rate, cm3/s; hi and wi are cyclone inlet height and width respectively, cm; and do is the gas outlet diameter, cm. This equation is in the same form as that proposed by Shepherd and Lapple [Ind. Eng. Chem., 31, 1246 (1940)] but gives only 37 percent as much pressure drop.

Liquid cyclone efficiency can be improved somewhat by introducing a coarse spray of liquid in the cyclone inlet. Large droplets which are easily collected collide with finer particles as they sweep the gas stream in their travel to the wall. (See subsection "Wet Scrubbers" regarding optimum spray size.) Cyclones may also be operated wet to improve their operation on dry dust. Efficiency can be improved through reduction in entrainment losses since the dust particles become trapped in the water film. Collision between droplets and dust particles aids collection, and adequate irrigation can eliminate problems of wall buildup and fouling. The most effective operation is obtained by spraying countercurrently to the gas flow in the cyclone inlet duct at liquid rates of 0.7 to 2.0 L/m3 of gas. There are also many proprietary designs of liquid separators using centrifugal force, some of which are illustrated in Fig. 14-109. Many of these were originally

Inlet

Inlet

Anderson Type Entrainment Separators

FIG. 14-109 Typical separators using impingement in addition to centrifugal force. (a) Hi-eF purifier. (V D. Anderson Co.) (b) Flick separator. (Wurster & Sanger, Inc.) (c) Type RA line separator. (Centrifix Corp., Bull. 220.)

FIG. 14-109 Typical separators using impingement in addition to centrifugal force. (a) Hi-eF purifier. (V D. Anderson Co.) (b) Flick separator. (Wurster & Sanger, Inc.) (c) Type RA line separator. (Centrifix Corp., Bull. 220.)

developed as steam separators to remove entrained condensate. In some designs, impingement on swirl baffles aids separation.

Impingement Separation Impingement separation employs direct impact and inertial forces between particles, the gas streamlines, and target bodies to provide capture. The mechanism is discussed in Sec. 17 under "Gas-Solids Separations." With liquids, droplet coalescence occurs on the target surface, and provision must be made for drainage without reentrainment. Calvert (R-12) has studied droplet collection by impingement on targets consisting of banks of tubes, zigzag baffles, and packed and mesh beds. Figure 14-110 illustrates some other types of impingement-separator designs.

In its simplest form, an impingement separator may be nothing more than a target placed in front of a flow channel such as a disk at the end of a tube. To improve collection efficiency, the gas velocity may be increased by forming the end into a nozzle (Fig. 14-110a). Particle collection as a function of size may be estimated by using the target-efficiency correlation in Fig. 17-39. Since target efficiency will be low for systems with separation numbers below 5 to 10 (small particles, low gas velocities), the mist will frequently be subjected to a number oftargets in series as in Fig. 14-110c, d, and g.

The overall droplet penetration is the product of penetration for each set of targets in series. Obviously, for a distribution of particle sizes, an integration procedure is required to give overall collection efficiency. This target-efficiency method is suitable for predicting efficiency when the design effectively prevents the bypassing or short-circuiting of targets by the gas stream and provides adequate time to accelerate the liquid droplets to gas velocity. Katz (R-16) investigated a jet and target-plate entrainment separator design and found the pressure drop less than would be expected to supply the kinetic energy both for droplet acceleration and gas friction. An estimate based on his results indicates that the liquid particles on the average were being accelerated to only about 60 percent of the gas velocity. The largest droplets, which are the easiest to collect, will be accelerated less than the smaller particles. This factor has a leveling effect on collection efficiency as a function of particle size so that experimental results on such devices may not show as sharp a decrease in efficiency with particle size as predicted by calculation. Such results indicate that in many cases our lack of predicting ability results, not from imperfections in the theoretical treatment, but from our lack of knowledge of velocity distributions within the system.

Katz (R-16) also studied wave-plate impingement separators (Fig. 14-110b) made up of 90° formed arcs with an 11.1-mm (0.44-in) radius and a 3.8-mm (0.15-in) clearance between sheets. The pressure drop is a function of system geometry. The pressure drop for Katz's system and collection efficiency for seven waves are shown in Fig. 14-111. Katz used the Souders-Brown expression to define a design velocity for the gas between the waves:

K is 0.12 to give U in ms-1 (0.4 for ft/s), and p; and pg are liquid and gas densities in any consistent set of units. Katz found no change in efficiency at gas velocities from one-half to 3 times that given by the equation.

Calvert (R-12) investigated zigzag baffles of a design more like Fig. 14-110e. The baffles may have spaces between the changes in direction or be connected as shown. He found close to 100 per collection for water droplets of 10 |m and larger. Some designs had high efficiencies down to 5 or 8 |m. Desirable gas velocities were 2 to 3.5 m/s (6.6 to 11.5 ft/s), with a pressure drop for a six-pass baffle of 2 to 2.5 cm (0.8 to 1.0 in) of water. On the basis of turbulent mixing, an equation was developed for predicting primary collection efficiency as a function of particle size and collector geometry:

where n is the fractional primary collection efficiency; ute is the drop terminal centrifugal velocity in the normal direction, cm/s; U„ is the superficial gas velocity, cm/s; n is the number of rows of baffles or bends; 8 is the angle of inclination of the baffle to the flow path, °; W is the width of the baffle, cm; and b is the spacing between baffles in the same row, cm. For conditions of low Reynolds number (NReD < 0.1) where Stokes' law applies, Calvert obtains the value for drop terminal centrifugal velocity of ute = dp ppa/18 |g, where dp and pp are the drop particle diameter, cm, and particle density, g/cm3, respectively; |g is the gas viscosity, P; and a is the acceleration due to centrifugal force. It is defined by the equation a = 2Ugj sin 8/W cos3 8. For situations in which Stokes' law does not apply, Calvert recommends substitution in the derivation of Eq. (14-227) for u of drag coefficients from drag-coefficient data of Foust et al. (Principles of Unit Operations, Toppan Co., Tokyo, 1959).

Calvert found that reentrainment from the baffles was affected by the gas velocity, the liquid-to-gas ratio, and the orientation of the baffles. Horizontal gas flow past vertical baffles provided the best drainage and lowest reentrainment. Safe operating regions with vertical baffles are shown in Fig. 14-112. Horizontal baffles gave the poorest drainage and the highest reentrainment, with inclined baffles intermediate in performance. Equation (14-228), developed by Calvert, predicts pressure drop across zigzag baffles. The indicated summation must be made over the number of rows of baffles present.

AP is the pressure drop, cm of water; pg is the gas density, g/cm3; Ap is the total projected area of an entire row of baffles in the direction of inlet gas flow, cm2; and At is the duct cross-sectional area, cm2. The

Nozzle

J Target

Gas flow

Nozzle

J Target

Gas flow

Centrifuge Separation Gas Liquid

FIG. 14-110 Typical impingement separators. (a) Jet impactor. (b) Wave plate. (c) Staggered channels. (Blaw-Knox Food & Chemical Equipment, Inc.) (d) Vane-type mist extractor. (Maloney-Crawford Tank and Mfg. Co.) (e) Peerless line separator. (Peerless Mfg. Co.) (f) Strong separator. (Strong Carlisle and Hammond.) (g) Karbate line separator. (Union Carbide Corporation) (h) Type E horizontal separator. (Wright-Austin Co.) (i) PL separator. (Ingersoll Rand.) j) Wire-mesh demister. (Otto H. York Co.)

FIG. 14-110 Typical impingement separators. (a) Jet impactor. (b) Wave plate. (c) Staggered channels. (Blaw-Knox Food & Chemical Equipment, Inc.) (d) Vane-type mist extractor. (Maloney-Crawford Tank and Mfg. Co.) (e) Peerless line separator. (Peerless Mfg. Co.) (f) Strong separator. (Strong Carlisle and Hammond.) (g) Karbate line separator. (Union Carbide Corporation) (h) Type E horizontal separator. (Wright-Austin Co.) (i) PL separator. (Ingersoll Rand.) j) Wire-mesh demister. (Otto H. York Co.)

value fD is a drag coefficient for gas flow past inclined flat plates taken from Fig. 14-113, while Ug is the actual gas velocity, cm/s, which is related to the superficial gas velocity U„ by Ug = Ug /cos 8. It must be noted that the angle of incidence 8 for the second and successive rows of baffles is twice the angle of incidence for the first row. Most of

Calvert's work was with 30° baffles, but the method correlates well with other data on 45° baffles.

The Karbate line separator (Fig. 14-110g) is composed of several layers of teardrop-shaped target rods of Karbate. A design flow constant K in Eq. (14-226) of 0.035 m/s (1.0 ft/s) is recommended by the

14-111 Pressure drop and collection efficiency of a wave-plate separator. (a) Pressure drop. (b) Efficiency : clearance between sheets. (Katz, M.S. thesis, Pennsylvania State University, 1958. )

2000 4000 6000 10,000 Reynolds number = D^Mp/fJ.

Was this article helpful?

0 0

Post a comment