426 Fermentation and Biochemical engineering handbo 5.0 PROCESS CONSIDERATIONS 5.1 Design Factors The engineer designing an ion exchange column operation usually will prefer to work with the simplest kinetic model and linear driving force approximations. The weakness of this approach is that any driving force law only regards the momentary exchange rate as a function of the solute concentration in the bulk solution and the average concentration in the article, neglecting the effect of concentration profiles in the particle Nevertheless, the linear driving force approach provides an approximation that is sufficiently accurate for the engineer 5.2 Scaling-up Fixed Bed Operations Rodrigues[ 66 has presented empirical and semi-empirical approaches In feedstream is co and the flow-rate is uo. The breakthrough point is usually set at the point where the effluent concentration increases to 5%of co. The design equations relate the total equilibrium ion exchange capacity(@)tothe volume of resin required()to the time of breakthrough(tB) In the empirical approach, the overall mass balance is given by the Eq(15) v=c0t/(1+8)Q Eq(16) =τ(1+8 (the stoichiometric time) Eq(17) (1-a)Q/E go (the mass capacity factor) Eq(18) τ=EW/a0 and v is the bed volume with void space e It is usually necessary to modify this resin amount by a safety fact (1. 2 to 1.5)to adjust for the portion of the total equilibrium capacity that can actually be used at flow rate u and to adjust for any dispersive effects that might occur during operation
426 Fermentation and Biochemical Engineering Handbook 5.0 PROCESS CONSIDERATIONS 5.1 Design Factors The engineer designing an ion exchange column operation usually will prefer to work with the simplest kinetic model and linear driving force approximations. The weakness ofthis approach is that any driving force law only regards the momentary exchange rate as a hnction of the solute concentration in the bulk solution and the average concentration in the particle, neglecting the effect of concentration profiles in the particle. Nevertheless, the linear driving force approach provides an approximation that is sufficiently accurate for the engineer. 5.2 Scaling-up Fixed Bed Operations Rodrigues[66] has presented empirical and semi-empirical approaches which may be used to design ion exchange columns when the solute in the feedstream is co and the flow-rate is uo. The breakthrough point is usually set at the point where the effluent concentration increases to 5% of co. The design equations relate the total equilibrium ion exchange capacity (e) to the volume of resin required (V,) to the time of breakthrough (b). In the empirical approach, the overall mass balance is given by the equation: where Eq. (16) t, = 5 (1 + 5) (the stoichiometric time) Eq. (17) 5 = (1- &)e/& Qo (the mass capacity factor) Eq. (18) 5 = &VhO (the space time) and Vis the bed volume with void space E. It is usually necessary to modify this resin amount by a safety factor (1.2 to 1.5) to adjust for the portion of the total equilibrium capacity that can actually be used at flow rate u and to adjust for any dispersive effects that might occur during operation
Ton Exchange 427 The semi-empirical approach involves the use of the mass transfer zones. This approach has been described in detail specifically for ion exchange resins by Passino. [67 He referred to the method as the operating line andregenerating line process design and used a graphical description to solve the mass transfer problems For the removal of Cat+ from a feedstream, the mass transfer can be modeled using Fig. 21. The upper part shows an element of ion exchange column containing a volume v of resin to which is added a volume Ver of the feedstream containing Ca**. It is added at a flow rate(Fi) for an exhaustion time tex. The concentration of Ca* as it passes through the column element is reduced from Mexi to rex2. Therefore, the resin, which has an equilibrium ion exchange capacity C, increases its concentration of Ca"?? to yexl In this model, fresh resin elements are continuously available at a flow rate (F)=v/o, which is another way of saying the mass transfer zone passes down through the column The lower part of Fig. 21 shows the operating lines for this process The ion exchange equilibrium line describes the selectivity in terms of a Freundlich, Langmuir or other appropriate model The equations for the points in the lower part are given by Eq (19) rex1=ro Ca* in the feedstream) Eq(20) yerl =yes2+(erl -xex2)over Eq(21) X (average Ca in the effluent) Eq(22) (Ca in the regenerated resin) and the slope of the operating line D : ver2 CF
Ion Exchange 427 The semi-empirical approach involves the use of the mass transfer zones. This approach has been described in detail specifically for ion exchange resins by Passin0.[~'1 He referred to the method as the operating line and regenerating line process design and used a graphical description to solve the mass transfer problems. For the removal of Ca" from a feedstream, the mass transfer can be modeled using Fig. 21. The upper part shows an element of ion exchange column containing a volume v of resin to which is added a volume V, of the feedstream containing Ca". It is added at a flow rate (Q) for an exhaustion time t, . The concentration of Ca* as it passes through the column element is reduced from xal to x,,. Therefore, the resin, which has an equilibrium ion exchange capacity Cy increases its concentration of Catt fromy,, to yal. In this model, fresh resin elements are continuously available at a flow rate (F') = v/t, , which is another way of saying the mass transfer zone passes down through the column. The lower part of Fig. 21 shows the operating lines for this process. The ion exchange equilibrium line describes the selectivity in terms of a Freundlich, Langmuir or other appropriate model. The equations for the points in the lower part are given by: Eq. (19) Xexl =xo (Ca" in the feedstream) (Ca" in the exhausted streamstream) x dv x,, (average Ca* in the effluent) = - v, Eq. (21) Eq. (22) Yd=O (Ca" in the regenerated resin) and the slope of the operating line:
428 Fermentation and Biochemical engineering Handb N0】 F N。】 88 色5
428 Fermentation and Biochemical Engineering Handbook .- 5 ti; C Q) 01 E -
Ton Exchange 429 The value of co, C and v are known so that for any Ver value, the slope of the operating line can be calculated from Eq. 23. The specific points: ro is given, xen is obtained by graphicintegration from the breakthrough curves After operation and regeneration, the value of yer? may not be zero but may be between 0.02 and 0.05 if the regeneration is not complete. The application of this technique has been described in terms of basic design parameters such as number of transfer units, the height of a transfer unit and mass transfer fficient. [65 The data generated with the laboratory column may be scaled-up to commercial size equipment. Using the same flow rate(on a mass basis)as used in the laboratory experiments, the appropriate increase in column size over that used in the laboratory is a direct ratio of the volumes to be treated compared to that treated in the laboratory equipment If a reasonable height to diameter ratio(approximately 1: 1)is obtained in the scaleup using the bed depth involved in the laboratory procedure, then that bed depth is maintained and the cross-sectional area of the column is increased. However, if the sizing is such that the column is much larger diameter than the bed depth, scaleup should be done to maintain a height to diameter ratio of approximately 1. The required resin volume is determined by maintaining the san flow conditions(liters of feed solution minute per cubic meter of installed resin) as was used in the laboratory operation Appropriate tank space must be left to accommodate the backwash operation. This is typically 50% of bed depth for cation exchange resins and 100% expansion in the case of anion exchange resins 5.3 Sample Calculation The purification of lysine-HCl from a fermentation broth will be used to illustrate the calculations involved in scaling-up laboratory data The laboratory fermentation broth, which is similar to the commercial contained 2.0 g/0. 1 l lysine, much smaller amounts of Ca", K and amino acids. The broth was passed through 500 ml of strong acid cation resin, Dowex(R HCR-S, in the NH4 form. The flow rate was 9 ml/min 1.77 mI min per cm of resin. It was determined that the resin capacity averaged 115 g of lysine-HCl per liter of resin. It may be noted that since the equivalent molecular weight of lysine-HCl is 109.6 g and the theoretical capacity of Dowex(B HCR-S is 2.0 meq/ml, the operating capacity is 52%of eoretical capacit
Ion Exchange 429 The value of c,,, C and v are known so that for any V, value, the slope of the operating line can be calculated from Eq. 23. The specific points: x, is given, x, is obtained by graphic integration fromthe breakthrough curves. After operation and regeneration, the value ofy,, may not be zero but may be between 0.02 and 0.05 ifthe regeneration is not complete. The application of this technique has been described in terms of basic design parameters such as number of transfer units, the height of a transfer unit and mass transfer coefficient. [as1[691 The data generated with the laboratory column may be scaled-up to commercial size equipment. Using the same flow rate (on a mass basis) as used in the laboratory experiments, the appropriate increase in column size over that used in the laboratory is a direct ratio of the volumes to be treated compared to that treated in the laboratory equipment. If a reasonable height to diameter ratio (approximately 1 : 1) is obtained in the scaleup using the bed depth involved in the laboratory procedure, then that bed depth is maintained and the cross-sectional area of the column is increased. However, if the sizing is such that the column is much larger in diameter than the bed depth, scaleup should be done to maintain a height to diameter ratio of approximately 1. The required resin volume is determined by maintaining the same mass flow conditions (liters of feed solution per minute per cubic meter of installed resin) as was used in the laboratory operation. Appropriate tank space must be left to accommodate the backwash operation. This is typically 50% of bed depth for cation exchange resins and 100% expansion in the case of anion exchange resins. 5.3 Sample Calculation The purification of lysine-HC1 from a fermentation broth will be used to illustrate the calculations involved in scaling-up laboratory data. The laboratory fermentation broth, which is similar to the commercial broth, contained 2.0 g/O. 1 1 lysine, much smaller amounts of Caw, Kf and other amino acids. The broth was passed through 500 ml of strong acid cation resin, DowexB HCR-S, in the NH,’ form. The flow rate was 9 ml/min or 1.77 mVmin per cm2 of resin. It was determined that the resin capacity averaged 1 15 g of lysine-HC1 per liter of resin. It may be noted that since the equivalent molecular weight of lysine-HC1 is 109.6 g and the “theoretical” capacity of DowexB HCR-S is 2.0 meq/ml, the operating capacity is 52% of theoretical capacity
430 Fermentation and Biochemical Engineering Handbook The commercial operation must be capable of producing 9,000 metric tons of lysine(as lysine-dihydrochloride-H2O) per year. With a 2.0 g/0.1 I concentration oflysine in the fermentation broth, the number of liters of broth to be treated each year are 9, 000 m tons 146.19(MW of lysine) 237. 12(MW ofL-HCl-H20 Eq,(24) x011×108=217×10/yx If the plant operates 85% of the time, the flow rate would have to be 27.7×1071x-1 0.85365days I day =3.73×1041/hr 24 hr avaii At a resin capacity of 115 g/l of resin, the amount of resin that must be I(resin) 219.12(MW of L-HCI.2.0 g 115g 146.19( MWof L)0.11 Eq(26) 3.73×104l/hr 9.71×103l/hr hr To obtain the maximum utilization of the resin in this operation, series bed operation(Carrousel)operation is recommended. This operation uses three beds of resin in a method having two beds operating in series while the product is being eluted from the third. The freshly regenerated resin is placed in the polishing position when the totally loaded lead bed is removed for
430 Fermentation and Biochemical Engineering Handbook The commercial operation must be capable of producing 9,000 metric tons of lysine (as lysinedihydrochloride-H,O) per year. With a 2.0 g/O. 1 1 concentration of lysine in the fermentation broth, the number of liters of broth to be treated each year are: 9,000 m tons 146.19 (MW of lysine) X Yr 237.12 (MWofL-HCl-H,O) x- Os1 x- lo6g = 27.7 x 107 i/p 2.0g Mton If the plant operates 85% of the time, the flow rate would have to be: 27.7 x lo7 l/yrxlx 1Yr 0.85 365 days 1 day 24 hr -=3.73x104 l/hr At a resin capacity of 1 15 g/1 of resin, the amount of resin that must be available is: l(resin) 219.12(MWof L-HCl) x-x 2.0g 115 g 146.19(MWof L) 0.1 1 3’73x1041/hr =9.71~103 l/hr hr To obtain the maximum utilization of the resin in this operation, series bed operation (Carrousel) operation is recommended. This operation uses three beds of resin in a method having two beds operating in series while the product is being eluted from the third. The freshly regenerated resin is placed in the polishing position when the totally loaded lead bed is removed for regeneration
on Erchange The elution/regeneration step, which includes backwashing, eluting, and rinsing the resin, might take up to four hours. Therefore, enough resin must be supplied to take up the lysine-HCl presented during that time. The resin requirement for the commercial scaleup operation would be Ea.(27 971×1031(resc+ed3.88×104l( resin))/bed Thus, three beds of 39 m resin each are required to produce 9,000 metric tons of lysine/ye 5.4 Comparison of Packed and Fluidized Beds Belter and co-workers!70] developed a periodic countercurrent proces for treating a fermentation broth to recover novobiocin. They found that they were able to scaleup the laboratory results to production operations if the two systems have similar mixing patterns and the same distribution of residence times in the respective columns. The mixing patterns are the same when the ace velocity(F/e)and the volume ration(VR/c)arethe same. This is she in Fig. 22 for the effluent concentration of novobiocin from laboratory and roduction columns The scaling-up of packed beds is subject to the difficulties of maintain ing even flow distributions. Removal of solution through screens on side walls is not recommended and the flow of resin from one section into another of much greater area could distort the resin flow profile The problems of scaling-up fluidized bed operations are more difficult in terms of design calculations, but flow distribution is more easily designed because of the mobility of the resin. The degree of axial mixing of the liquid and the resin has to be taken into account when calculating the changes necessary in the bed diameter and bed height. Figure 23 shows the increase in bed height necessary when scaling-up packed and fluidized beds with bed diameter increases. [711 Mass transfer coefficients have been correlated for packed and fluid- ized beds. 12) The mass transfer coefficients for packed beds are 50 to 100% greater than for fluidized bed The volume ofresin in a packed bed is about half that in a fluidized bed but the packed bed column may be up to eight times smaller. Despite this, complete fluidized bed operation may still be smaller than a fixed bed operation. Also, fluidized bed columns do not operate at high pressures so they can be constructed more economically
Ion Exchange 431 The elutionhegeneration step, which includes backwashing, eluting, and rinsing the resin, might take up to four hours. Therefore, enough resin must be supplied to take up the lysine-HC1 presented during that time. The resin requirement for the commercial scaleup operation would be: 9.71 x lo3 l(resin) 4 hr hr bed Eq. x-=3.88x104 l(resin)/bed (27) Thus, three beds of 39 m3 resin each are required to produce 9,000 metric tons of lysine/year. 5.4 Comparison of Packed and Fluidized Beds Belter and co-w~rkers[~~] developed a periodic countercurrent process for treating a fermentation broth to recover novobiocin. They found that they were able to scaleup the laboratory results to production operations ifthe two systems have similar mixing patterns and the same distribution of residence times in the respective columns. The mixing patterns are the same when the space velocity (FN,) and the volume ration (&IC) are the same. This is shown in Fig. 22 for the effluent concentration of novobiocin from laboratory and production columns. The scaling-up of packed beds is subject to the difficulties of maintaining even flow distributions. Removal of solution through screens on side walls is not recommended and the flow of resin from one section into another of much greater area could distort the resin flow profile. The problems of scaling-up fluidized bed operations are more difficult in terms of design calculations, but flow distribution is more easily designed because of the mobility of the resin. The degree of axial mixing of the liquid and the resin has to be taken into account when calculating the changes necessary in the bed diameter and bed height. Figure 23 shows the increases in bed height necessary when scaling-up packed and fluidized beds with bed diameter Mass transfer coefficients have been correlated for packed and fluidized beds.[72] The mass transfer coefficients for packed beds are 50 to 100% greater than for fluidized beds. The volume of resin in a packed bed is about halfthat in a fluidized bed, but the packed bed column may be up to eight times smaller. Despite this, a complete fluidized bed operation may still be smaller than a fixed bed operation. Also, fluidized bed columns do not operate at high pressures so they can be constructed more economically
432 Fermentation and Biochemical Engineering Handbook d。omti"zin Figure 22. Comparison of experimental curves for laboratory and production columns. ED。 N CORRELATION PLU FLON COLUMN DIAMETER ( NETER) Figure 23. Scaleup relationships for fluidized beds. [71)
432 Fermentation and Biochemical Engineering Handbook 01 ‘1 1’ I’ ” 0 30 60 90 120 150 110 210 240 210 TIME IN HlNUTfS Figure 22. Comparison of experimental curves for laboratory and production column^.^^^] Figure 23. Scaleup relationships for fluidized beds.[7’]
Ion Exchange 433 5.5 Chromatographic Scale-Up Procedures The aim of scaling-up a chromatographic process is to obtain the same yield and product quality in the same time period on laboratory, preparative and industrial scale. Laboratory analytical purifications tend to be optimized only for resolution of individual solutes, however, at the preparative and production scale, it is necessary also to maximize throughput. Table 14173J shows the effect of chromatographic operating parameters on resolution and throughput. While column length is a critical factor for resolution with gel filtration and isocratic elutions, it has little effect on resolution with gradient elutionin adsorption chromatography. The wall effects on resolution are very noticeable with small radius columns, but decrease as the column length is increased Table 14. Effects of Process Parameters on Resolution and Throughput!31 Resolution Throughput Parameter varies with varies with Column Radius(r) Some effect T Positive effect Viscosity(n) Negative effect Sample Volume(F 1/V-Ve Flow Rate) Voser and Walliser!74 viewed scaleup as a three step process involving election of a process strategy, evaluation of the maximum and optimal bed height and finally column design. The selection of a process strategy involves choosing the direction of flow, frequency of backwashing, the operation with positive head pressure or only hydrostatic pressure, and the use of column or a series of columns in batch or semi-continuous operation, The
Ion Exchange 433 5.5 Chromatographic Scale-up Procedures The aim of scaling-up a chromatographic process is to obtain the same yield and product quality in the same time period on laboratory, preparative and industrial scale, Laboratory analytical purifications tend to be optimized only for resolution of individual solutes, however, at the preparative and production scale, it is necessary also to maximize throughput. Table 14[731 shows the effect of chromatographic operating parameters on resolution and throughput. While column length is a critical factor for resolution with gel filtration and isocratic elutions, it has little effect on resolution with gradient elution in adsorption chromatography. The wall effects on resolution are very noticeable with small radius columns, but decrease as the column length is increased. Table 14. Effects of Process Parameters on Resolution and Throughput[73] Resolution Throughput Parameter varies with varies with Column Length (L) L I/L Column Radius (r) Some effect r2 Temperature (T) Positive effect T Viscosity (q) Negative effect 1h Sample Volume (V) 14 ‘-‘optimum I V Flow Rate (J) 1WJoptimuml J Voser and Walli~er[~~] viewed scaleup as a three step process involving selection of a process strategy, evaluation of the maximum and optimal bed height and finally column design. The selection ofa process strategy involves choosing the direction of flow, frequency of backwashing, the operation with positive head pressure or only hydrostatic pressure, and the use of a single column or a series of columns in batch or semi-continuous operation. The
34 Fermentation and Biochemical Engineering Handbook maximum feasible bed height is determined by keeping the optimal labora tory-scale specific volume velocity(bed volume/hour)constant. The limiting factors will be either pressure drop or unfavorable adsorption/desorption kinetics since the linear velocity also increases with increasing bed height Bed heights from 15 cm to as high as 12 m have been reported. The column diameter is then selected to give the required bed volume. The column design combines these column dimensions with the practical considerations of available space, needed flexibility, construction difficulty and flow distribu tion and dilution for the columns The scaleup considerations of column chromatography for protein isolation has been described by Charmand Matteo. [75] When several hundred liters of a protein feedstream must be treated the resin may be suspended in the solution, removed after equilibration by filtration and loaded into a column from which the desired proteins may be eluted. Adsorption onto a previously packed column was not recommended by them since they feared suspended particles would clog the interstices of the column, causing reduced flow rates and increased pressure drops across the column. The reduced flow rate may lead to loss of enzyme activity because of the increased time the protein is adsorbed on the resin is important that all of the resin slurry be added to the column in one operation to obtain uniform packing and to avoid the formation of air pockets in the column. An acceptable altermative, described by Whatman, b allows the addition of the adsorbent slurry in increments. when the resin has settled to a packed bed of approximately 5 cm, the outlet is opened. The next increment of the slurry is added after the liquid level in the column has dropped. It is important that the suspended adsorbent particles do not completely settle between each addition Stacey and coworkers! 77 have used the relationships shown in Table 15 to scaleup the purification from an 8 mg protein sample to a 400 mg sample. The adsorbent used in both columns was a Delta-Pak wide-pore c- 18 material. When eluting the protein, the flow rate should change so that the linear velocity ofthe solvent through the column stays the same. The flow rate is proportional to the cross-sectional area of the column. The gradient duration must be adjusted so that the total number of column volumes delivered during the gradient remain the same. As with size exclusion chromatography, the mass load on the preparative column is proportional to the ratio of the column volumes. Figure 24 shows that the chromatograms from the 8 mg separation is very similar to that obtained for the 400 mg
434 Fermentation and Biochemical Engineering Handbook maximum feasible bed height is determined by keeping the optimal laboratory-scale specific volume velocity (bed volumehour) constant. The limiting factors will be either pressure drop or unfavorable adsorptioddesorption kinetics since the linear velocity also increases with increasing bed height. Bed heights from 15 cm to as high as 12 m have been reported. The column diameter is then selected to give the required bed volume. The column design combines these column dimensions with the practical considerations of available space, needed flexibility, construction difficulty and flow distribution and dilution for the columns. The scaleup considerations of column chromatography for protein isolation has been described by Charm and Matte0.1~~1 When several hundred liters of a protein feedstream must be treated, the resin may be suspended in the solution, removed after equilibration by filtration and loaded into a column from which the desired proteins may be eluted. Adsorption onto a previously packed column was not recommended by them since they feared suspended particles would clog the interstices of the column, causing reduced flow rates and increased pressure drops across the column. The reduced flow rate may lead to loss of enzyme activity because of the increased time the protein is adsorbed on the resin. It is important that all of the resin slurry be added to the column in one operation to obtain uniform packing and to avoid the formation of air pockets in the column. An acceptable alternative, described by Whatman,[76] allows the addition ofthe adsorbent slurry in increments. When the resin has settled to a packed bed of approximately 5 cm, the outlet is opened. The next increment of the slurry is added after the liquid level in the column has dropped. It is important that the suspended adsorbent particles do not completely settle between each addition. Stacey and coworkers[77] have used the relationships shown in Table 15 to scaleup the purification from an 8 mg protein sample to a 400 mg sample. The adsorbent used in both columns was a Delta-Pak wide-pore C- 18 material. When eluting the protein, the flow rate should change so that the linear velocity ofthe solvent through the column stays the same. The flow rate is proportional to the cross-sectional area of the column. The gradient duration must be adjusted so that the total number of column volumes delivered during the gradient remain the same. As with size exclusion chromatography, the mass load on the preparative column is proportional to the ratio of the column volumes. Figure 24 shows that the chromatograms from the 8 mg separation is very similar to that obtained for the 400 mg sample
lon Exchange 435 Table 15 Scale-Up Calculations[77 Small Scale Preparative Scale Column dimensions 0.39×30cm 3×25cm Flow Rate Scale Factor (3 (0.39)2=59 1.5 ml/min 90 mi/min Sample load Scale Factor (30)2×25 (0.39)2×25 Gradient Duration Calculation 40 min 1.5×40 Gradient Duration)-167 (1.5)2xx25 Grad. duration 33 min 3.0 8mg sample 1.5 1520 Minutes Figure 24. Laboratory and preparative scale separation of cytochrome C digest. 771
Ion Exchange 435 .4 . Table 15 Scale-up Calculations[77] 8mg Sample Small Scale Preparative Scale Column Dimensions 0.39 x 30 cm 3 x 25 cm Flow Rate Scale Factor (3.0)* = 59 (0.39)2 Sample Load Scale Factor = 49 (3.0)2 x 25 (0.39)2 x 25 1.5 mVmin 90 ml/min 400 mg Gradient Duration Calculation 40 min 33 min column 90 x (Gradient Duration) - 16.7 1'5x40 46.7 - (0.195)2 zx 30 Volume (1.5)2lcx25 Grad. Duration = 33 min. 0.8 f 1 O" 0 8 Minutes Figure 24. Laboratory and preparative scale separation of cytochrome C digest.[77]