Gas-Liquid Mass Transfer Using Surface-Aeration in Stirred Vessels, with Dual Impellers

Gas-Liquid Mass Transfer Using Surface-Aeration in Stirred Vessels, with Dual Impellers

0263±8762/00/$10.00+0.00 q Institution of Chemical Engineers Trans IChemE, Vol 78, Part A, April 2000 GAS-LIQUID MASS TRANSFER USING SURFACE-AERATION...

269KB Sizes 0 Downloads 26 Views

0263±8762/00/$10.00+0.00 q Institution of Chemical Engineers Trans IChemE, Vol 78, Part A, April 2000

GAS-LIQUID MASS TRANSFER USING SURFACE-AERATION IN STIRRED VESSELS, WITH DUAL IMPELLERS P. C. LINES (ASSOCIATE MEMBER) Zeneca Agrochemicals, Hudders®eld, UK

M

ass transfer is calculated, for air-water systems, using measurements from a dissolved oxygen probe with a simple model. Most measurements are taken at the 20 litre scale, but successful mass transfer measurements have also been made at the 5000 litre scale. The full-scale designs are illustrated for a number of different processes, and examples given of the reduction in cycle-time achievable by a change of impeller con®guration. The effects of geometry and gassed-power input on the mass transfer coef®cient are also discussed. Keywords: mass transfer; agitation; surface aeration; gassed-power input

INTRODUCTION

A number of papers in the literature refer to gas entrainment, or incorporation from a liquid surface, and these are usefully reviewed by Patwardhan and Joshi1, plus Forrester et al.2. However, most of the references quoted in these two papers have mechanically-complicated designs, and/or do not discuss mass transfer results. Speci®cally, most of the papers discuss systems which involve the use of draught-tubes, multiple-shafts or hollow-shaft designs. None of these systems are considered to be suitable for ¯exible ®ne chemical process vessels, because;

Gas-liquid mass transfer is a common industrial process in a number of industries and is widely reported in the literature. In most cases, the gas is fed into the impeller using a spargering or dip-leg, and the feed-rate is usually controlled by some yield or concentration measurement. However, in a number of cases, particularly semi-batch operations, the gas uptake decreases markedly as the reaction progresses, either because of a reduction in the reactant concentration, and/or because the reaction mechanism has changed from mass transfer to chemical-rate control. In such cases, the cycle time of the stage is often dominated by a `tail’ of the ®nal 10-20% conversion. This type of operation can be typi®ed by a `dead-end hydrogenation’ (for example), and has been referred to in the literature by Oldshue and Wisniak3,4. In effect, the agitator design has to disperse gas from a pipe or ring sparger at quite high ¯ow-rates, draw gas down from the vapour space towards the end of the gas-liquid reaction, and perform other duties as required (eg. heat-transfer, solids suspension, etc.). There are other references to this type of system geometry in the literature, but most refer to conventional gas-feed arrangements, `tall’ batches and/or dif®cult rheology. There are a number of gas-liquid reactions in industry, which have long cycle times, due to a progressive reduction in the mass transfer coef®cient, and/or a transfer of overall control from mass transfer to chemical rate. These reactions are usually batch or semi-batch processes, at small to moderate scale, and often involve high-value materials. The starting point for the current experimental trials was the need to improve a batch catalytic hydrogenation, with a cycle-time of about 12 hours, at the 7 cubic metre scale. Any new design had to work ®rst time, be mechanically simple but ¯exible, maintain high levels of safety, and have low costs. These constraints virtually rule out multiple-shaft designs, external-loop reactors, draught-tubes, and special impeller designs. In order to arrive at a suitable design, a number of dual-impeller combinations were tried, in conjunction with a number of different baf¯e con®gurations.

(a) they are mostly characterized for surface incorporation only, or, (b) they would be dif®cult or expensive to incorporate into plant vessels or, (c) there is limited design and scale-up information. In the papers reviewed in References 1 and 2, there is no general consensus about design correlations, even when the geometry is the same. Although there is some reported agreement on the relationship between surface incorporation and other parameters, there are no design methods nor scale-up rules. In general it might be argued that scale-up should be between constant power-per-unit-volume and constant Froude Number. The design, for the trials reported here, is based on the traditional agitation systems for `dead-end’ hydrogenations, particularly in the vegetable-oil industry, and as described by Oldshue3 and Wisniak et al.4, for example. This design is usually an angle-blade turbine and discturbine, on the same shaft, in a baf¯ed vessel. The best baf¯e con®guration is reported3 to be four half-height wall baf¯es.

EXPERIMENTAL Most of these experimental trials have been carried out in a ¯at-bottom 304 mm internal diameter `Perspex’ vessel, using air and tap water under ambient conditions. A sketch of the rig is shown in Figure 1. The procedure is to precalibrate the instruments, and then to set up the impeller and baf¯e geometry, for each trial. Water is added to the vessel, 342

GAS-LIQUID MASS TRANSFER USING SURFACE AERATION IN STIRRED VESSELS

343

Figure 2

Instrumentation

Figure 1

and the agitator set to rotate at about 50% of the trial speed; this initial speed is to allow de-aeration by sparging, without aeration from the surface. De-aeration of the water is achieved by bubbling nitrogen gas through a sparge ring, at about 1 vvm. When the water has been de-aerated, the nitrogen ¯ow is stopped, the agitator speed is increased to test speed, and data collection is started. A test run is terminated once the dissolved oxygen concentration has reached a maximum and level value. This procedure is then repeated for each different set of conditions. The vessel is pre-calibrated for volume and batch level, so that the level is accurately repeatable by volume additions. Essentially the same procedure is used for the 5000 litre-scale trials, but it proved to be much more dif®cult to de-aerate the water, prior to a test run.

Shaft Torque Impeller torque is measured by a strain-gauge, installed on an agitator half-shaft. This strain-gauge unit has a range of 5 Nm, and is manufactured to our own design, by Astech Electronics Ltd. of Binsted, Hampshire, UK. Calibration is carried out before each experimental campaign, using a balanced-beam and accurate balance-weights. Calibration of the strain-gauge gives a conversion factor, in units of Nm/ volt. This is used to multiply the strain-gauge signal; giving torque in Nm, and this, in turn, can be multiplied by 2pN, to give the impeller power consumption in watts. The straingauge unit is attached directly to the agitator shaft, between the motor and impellers (see Figure 1), using sleeves. Data from the strain-gauge is transmitted, via a radio transmitterreceiver, to a signal ampli®er/conditioner. Output from this unit is recorded, with Unigen data-logger software, on a local PC. Dissolved-oxygen concentration A polarographic dissolved-oxygen probe is immersed in the batch ¯uid; either through the vessel wall or at the end of a dip-leg. This is a fast-response probe manufactured by Electrosence of Frodsham, UK. The response time of the probe is tested before each campaign, and must be less than 1.5 seconds, for a change from pure nitrogen to air. This is done simply, by immersing the probe tip completely in a

Table 1. Dimensions of impellers used for these trials. Impeller blade Diameter Ratio (D/T)

Number

Angle* (8 )

Width ratio (W/D)

Thickness ratio (x/W)

Angle-blade turbine Angle-blade turbine Angle-blade turbine Angle-blade turbine Angle-blade turbine

0.50 0.30 0.51 0.50 0.50

4 4 4 6 4

60 60 45 60 60

0.23 0.39 0.24 0.24 0.23

0.049 0.046 0.044 0.045 0.046

Concave-blade disc-turbine Concave-blade disc-turbine

0.30 0.36

6 6

90 90

0.33 0.22

0.078 0.0064

Flat-blade disc-turbine Flat-blade disc-turbine

0.51 0.37

6 6

90 90

0.19 0.20

0.040 0.067

Hydrofoil

0.50

3

30

variable

Impeller type

* impeller blade angle is measured to the horizontal.

Trans IChemE, Vol 78, Part A, April 2000

,

variable

344

LINES Table 2. Principal dimensions of model baf¯es used in these trials. Dimentsional ratios Type

Number

Width (Wb/T)

Height (Hb/T)

Clearance (Zb/T)

Thickness (x/Wb)

4 4 2

0.089 0.075 0.122

1.32 0.336 2

[0] [0] 0.23

0.09 0.11 0.35

Full Half Beavertail

¯ow of pure nitrogen from a gas cylinder, then moving it away into free air. The position of the probe in the vessel is ®xed in relation to the position of the bottom impeller, throughout these trials. Agitator speed Shaft speed is measured using an infra-red tachometer, and a re¯ective strip, mounted on the agitator shaft. The precalibrated instrument is a RS205-536 model made by Radio Spares of Corby, Northamptonshire, UK. It has a certi®ed accuracy of better than 6 1 rpm over the range of interest. Impellers A variety of impeller types have been used, in a number of different combinations. A list of these impellers and their geometries is given in Table 1. Baf¯es Three sets of vessel baf¯es have been used: full baf¯es, half baf¯es and beavertail baf¯es. `Full baf¯es’ refer to 4 standard wall baf¯es; `half baf¯es’ refer to 4 standard wall baf¯es with reduced baf¯e height, and `beavertail baf¯es’ are based on a glass-lined-steel design, in the shape of a ¯attened tube. Details of the actual baf¯es used in these trials are given in Table 2. Data Handling The calculation of the overall gas-liquid mass transfer coef®cient kLa from the transient response curve, assumes that (i) the liquid phase is well mixed, (ii) gas absorption is liquid-phase controlled and

Figure 3

(iii) the oxygen concentration in the liquid at the interface is in equilibrium with the oxygen concentration in the gas. The rate equation is therefore: dC C) = kL a ´ (C ê dt kL a = ln

C ê C

(1)

C0 C

t (2) ê Thus a plot of the logarithm of the concentration-ratio term, against time, will give a slope of kLa, and, since the output of the dissolved-oxygen probe is directly proportional to the liquid oxygen concentration, then the concentration-ratio term in equation (2) can be expressed directly in probe output-voltages, or chart-recorder units. A typical data set, with a graphical calculation of the mass transfer coef®cient, is shown in Figure 2. Note that, at the 20 litre scale, the ¯uid mixing time is about 10 seconds and, the whole experiment lasts only about 2 minutes: this makes it vital that the oxygen-probe has a fast response. In general, the results are taken to be acceptable, if the R2 value for the slope of the graph is greater than 0.998. RESULTS AND DISCUSSION The initial trials use a combination of a down-pumping 4 ´ 458 angle-blade turbine (short-hand nomenclature: 4MFD45), mounted on the same shaft, above a 6 ¯at-blade `Rushton’ disc-turbine (short-hand nomenclature: 6RDT). A 30 litre vessel is used with 3 different baf¯e geometries, different batch volumes, and different agitator speeds. The effect of agitation speed on the gas-liquid mass

Figure 4

Trans IChemE, Vol 78, Part A, April 2000

GAS-LIQUID MASS TRANSFER USING SURFACE AERATION IN STIRRED VESSELS

345

Figure 7

Figure 5

transfer coef®cient, at constant H/T ratio, is shown in Figure 3. As might be expected, the mass transfer coef®cient increases as the agitation speed increases, although the effect is different for each set of baf¯es. The highest mass transfer coef®cients, for this agitation system, are always obtained with the 4 half-height wall baf¯es, as reported by Oldshue3. At speeds up to about 275 rpm, Beavertail baf¯es give better mass transfer than 4 full-height wall baf¯es (`standard’), but above that speed, there is little difference between the use of 2 Beavertail, and 4 standard wall baf¯es. These comments also apply to a comparison of power-perunit-volume, but only the full- versus half-height wall baf¯es results are clear-cut, due to errors of about + / 20%. ê A similar set of experimental trials, with the same agitator geometry, examine the effect of liquid level on the mass transfer coef®cient, at constant power/unit-volume. The results are displayed in Figure 4, and show that surface incorporation, as measured by the gas-liquid mass transfer

Figure 6

Trans IChemE, Vol 78, Part A, April 2000

coef®cient, decreases as the liquid level increases. In all cases, the performance of 4 standard wall baf¯es is poorest, as the liquid height ratio rises from 1.0 (22 litres) to 1.4 (24.5 litres). For the same increase in liquid height, mass transfer coef®cients for both the 2 Beavertail and the 4 halfheight baf¯es falls dramatically, but the fall is greater for the half-height baf¯es. At small increases in the liquid level, the half-height baf¯es give better gas incorporation than the 2 Beavertail system; however, above a liquid height ratio of about 1.1, the 2 Beavertail system mass transfer coef®cient is reduced by the smallest amount. Thus, if there is to be a large change in batch heights, 2 Beavertail baf¯es are preferred, for this agitator design. For most batch or semibatch processes, where the changes in batch heights are usually moderate-to-small, the half-height baf¯es are preferred. In general, however, the advantages of partial or reduced baf¯es over standard baf¯es virtually disappears, once the liquid height ratio reaches about 1.5: at these liquid levels, the surface aeration is very weak and eventually becomes minimal.

Figure 8

346

LINES

Figure 9

As a result of the small-scale trials, described brie¯y above, this agitation geometry was used to improve the performance of the hydrogenation stage of a ®ne-chemical intermediate. The `old’ process sparged hydrogen gas underneath a single disc-turbine, in a fully-baf¯ed vessel, with a batch capacity of about 7 cubic metres, and a hydrogenation time of about 12 hours (see Figure 5). This stage was the bottle-neck of the whole process, was dif®cult to control, and gave unexpectedly high levels of impurities; not seen during laboratory development work. The `new’ agitation system consists of a 4 ´ 458 angle-blade turbine mounted above a 6-blade disc-turbine on the same shaft. Although the impeller design and speed have been changed, the existing motor and gear-box have been retained, and the speci®c power was similar. The existing baf¯es were reduced to half-height wall baf¯es, and the whole system is sketched in outline in Figure 6. Replacing the existing single disc-turbine by a dual system, allows the uptake of about 35% more hydrogen, in a slightly lower cycle time (, 17%), and gives other bene®ts (see Figure 7), such as

better process control, and fewer by-products. In essence, the ®rst quarter of the cycle had better liquid-gas mixing (inferred subsequently by analysis of impurity pro®les). The main advantages of better temperature control, and shorter processing times, are seen in the last half of the cycle. Some of these detailed improvements are not seen in Figures 5 and 7, due to the reduced number of data points, which were chosen to simplify the graphs, and to remove more frequent oscillations. The outstanding problem with this particular process, is that the cycle-time is now governed by the rate of heat removal, in order to maintain the batch temperature below 308 C in this vessel. In a similar problem, with a different chemical intermediate, the heat removal was improved by the installation of helical coils in the vessel. In this case, the use of a nonpreferred geometry is determined by the interference from the inter-connection of the coils, dip-pipe and supports/ baf¯es. In essence, the various dip-legs and thermometer pockets restrict the diameter of the top impeller to less than 0.4T: the lower impeller was replaced with a concave-blade disc-turbine of the same diameter. The ®nal agitator design is shown in Figure 8; however, note that the system is effectively fully baf¯ed. Despite these problems, the use of an angle-blade turbine, mounted over a disc-turbine, has again led to substantial improvements in performance ( > 30% cycle-time reduction). This was with the same agitator motor, at similar power consumptions, and due mainly to a shorter, tail-end, cycle-time. This type of dual impeller agitation system has been successfully applied to hydrogenations, air oxidations, carboxylations and carbonylations. In all cases, there is no established rule for scale-up, from laboratory to plant. When an opportunity presented itself, the decision was therefore taken to do some large-scale surface-incorporation trials at Zeneca. This involved the purchase of a large-body oxygen probe (from Electrosence Ltd), which could be attached to a dip-leg in the plant vessel. The principal geometry of the plant is given in Figure 9: this is close to, but not exactly the same as a laboratory geometry (LINES(A) = 0.51T 4MFD45/0.36T 6CBDT; LINES(B) = 0.5T4MFD60 / 0.37T 6RDT). Although the same laboratory experimental process is used in the plant trials, practical problems mean that the large-scale mass transfer results are considerably lower than can be obtained in the laboratory, and under fully-operational conditions. When the data from these plant commissioning trials is compared with the laboratory results

Figure 10

Trans IChemE, Vol 78, Part A, April 2000

GAS-LIQUID MASS TRANSFER USING SURFACE AERATION IN STIRRED VESSELS

347

Figure 11

for similar geometries, then slightly different trends are observed (see Figure 10). The results, as displayed, indicate that, for this geometry, scale-up based on equal power-perunit-volume, is not fully applicable. At low to moderate power inputs, the laboratory results can over-predict plantscale mass transfer performance, based on scale-up at constant speci®c power input. Like most authors, we believe that scale-up at constant speci®c power input is the best method for sparged-gas systems. These dual impeller designs are seen as a ¯exible and relatively cheap new, or retro-®t systems, to overcome the speci®c problem of extended reaction times, caused by a reduction in the gas-liquid mass transfer coef®cient. It is not the most ef®cient design for surface aeration, as can be seen in Figure 11, where the performance of these designs is compared to other literature data, as listed by Patwardhan and Joshi1. These results are for 300 mm diameter vessels, using tap-water(apart from 1), H = T, and the sulphiteoxidation test method for mass transfer coef®cients. The system used by Matsumura et al.5 is seen to be the most ef®cient of these systems. However, the results of the laboratory trials, reported here, give similar or better ef®ciencies to the systems used by Albal et al.6, Fuchs et al.7, Takase et al.8 and Wu9, for example, as can be seen on the composite graph in Figure 11. Note that none of these references have demonstrated the ability to handle more conventional gas-liquid dispersion, with gas feed from a submerged dip-pipe or sparge-ring. CONCLUSIONS A geometrically-simple dual impeller system has been shown to give good gas-liquid performance in a number of industrial-scale reactors. The surface-incorporation system works for a number of baf¯e designs, and for different gasliquid systems. It can be retro-®tted in place of a single turbine; usually without having to change the motor or gear-box. NOMENCLATURE D C C*

impeller swept-diameter, m dissolved oxygen concentration [chart reading] at any time dissolved oxygen concentration at equilibrium [chart reading]

Trans IChemE, Vol 78, Part A, April 2000

C0 Hb kLa t T V vvm W Wb x Z Zb

dissolved oxygen concentration at the start of a run[chart reading] baf¯e height, m gas-liquid mass transfer coef®cient (based on liquid volume) s ±1 time, s vessel internal diameter, m ¯uid volume, m3 (volumetric ¯ow of gas/minute) / (volume of liquid), min ±1 impeller blade width, m baf¯e width, m thickness, m clearance from bottom of the impeller blades to vessel base, m clearance from bottom of a baf¯e to vessel base, m

REFERENCES 1. Patwardhan, A. W. and Joshi, J. B., 1998, Design of stirred vessels with gas entrained from free liquid surface, Can J Chem Eng 76: 339±364. 2. Forrester, S. E., Rielly, C. D. and Carpenter, K. J., 1998, Gas-inducing impeller design and performance characteristics, Chem Eng Sci 53 (4): 603±615. 3. Oldshue, J. Y., 1979, The role of mixing in hydrogenation processes, AIChE Meeting, Boston, USA. 4. Wisniak, J., Stefanovic, S., Ruben, E., Hoffman, Z. and Talmon, Y., 1971, Mixing effects in dead-end hydrogenation of oils: Sul®de oxidation, J Am Oil Chem Soc, 48: 379±383. 5. Matsumura, M., Sakuma, H., Yamagata, T. and Kobayashi, J., 1982, Performance of oxygen transfer in a new gas entraining fermenter, J Ferment Technol, 60 (5): 551±563. 6. Albal, R. S., Shah, Y. T., Schumpe, A. and Carr, N. L., 1983, Mass transfer in multiphase agitated contactors, Chem Eng J, 27: 61±80. 7. Fuchs, R., Ryu, D. D. Y. and Humphrey, A. E., 1971, Effect of surface aeration on scale-up procedures for fermentation processes, Ind Eng Chem Process Des Dev, 10: 190±196. 8. Takase, H., Unno, H. and Akehata, T., 1984, Oxygen transfer in a surface aeration tank with a square cross section, Int Chem Eng, 24: 128±134. 9. Wu, H., 1995, An issue on applications of a disk turbine for gas-liquid mass transfer, Chem Eng Sci, 50 (17): 2801±2811.

ADDRESS Correspondence concerning this paper should be addressed to Mr P. C. Lines, Zeneca Agrochemicals, Zeneca Hudders®eld Works, PO Box A38, Leeds Road, Hudders®eld, Yorkshire HD2 1FF, UK. (E-mail philip.lines@ agmanuk.zeneca.com). The manuscript was received 5 October 1999 and accepted for publication after revision 22 March 2000. The work was originally presented at the Fluid Mixing 6 Symposium held 7±8 July 1999 at the University of Bradford, UK.