2 BioPharm AUGUST 2002

Process Development

computer model of the entire process canprovide a common reference and anevaluation framework that facilitates processdevelopment. A model shows the effects ofprocess changes, and those changes can bereadily evaluated and systematicallydocumented.

A reliable model can pinpoint the mostcost-sensitive areas — the economic hotspots — of a complex process. Those spotsare usually capital intensive with highoperating cost or with low yields andproduction throughputs. The findings fromsuch analyses can focus additional lab orpilot-plant studies to optimize those processsteps. Experimenting on the computer withalternative process setups and operatingconditions allows a company to reducecostly and time-consuming laboratory andpilot plant efforts.

The environmental effect of a process canreadily be evaluated with computer models.Material balances calculated for theprojected large-scale manufacturing revealenvironmental hot spots. Those are usuallyprocess steps that require solvents orregulated materials with high disposal costs.Environmental issues not addressed duringprocess development can lead to seriousheadaches during manufacturing becauseafter a process is approved by regulatoryagencies, it is costly and time-consuming tomake process changes. That is particularlytrue for biopharmaceutical production, aboutwhich it is commonly said the processmakes the product.

Facility design and selection. When processdevelopment nears completion at the pilotlevel, simulation tools are used tosystematically design and optimize the large-scale process for commercial production.Good computer models can facilitate thetransfer of process technology and facilitiesdesign. If a new facility needs to be built,

Throughput Analysis and Debottlenecking of Biomanufacturing FacilitiesA Job for Process Simulators

Bottlenecks are everywhere, fromthe freeway overpass during themorning commute to the longlines at the supermarket. Butbottlenecks in a manufacturingprocess are bad for business.Computer models can help youeliminate those conditions orsituations that are retarding yourprogress. Whether the goal isstrategic planning, evaluatingalternatives, purchasingequipment, appraising a facility, oroptimizing production processes,simulation tools can improve youranalysis.

Process simulators and other modelingtools are gaining acceptance andpopularity in the biotech industry.Such tools are mainly used to evaluate“what-if” scenarios and to optimize

integrated processes. Tasks handled byprocess simulators include material andenergy balances of integrated processes,equipment sizing, cost analyses, schedulingof batch operations, environmental impactassessments, throughput analyses, anddebottlenecking (removing a condition orsituation that limits process throughput).Process simulation tools can be usedthroughout the life cycle of processdevelopment and product commercialization(Figure 1).

Using Process Simulation ToolsSimulation tools can be used at many stagesduring the commercialization process.

Idea generation. When product and processideas are first conceived, process simulatorscan be used for project screening andselection and for strategic planning based onpreliminary economic analyses.

Process development. During preclinical andclinical testing of a drug candidatecompound, a company’s processdevelopment group is looking into theoptions available for manufacturing,purifying, and characterizing the drugsubstance and for formulating it as a drugproduct. At this stage, the process undergoesmany changes. New synthetic routes areinvestigated: New recovery and purificationoptions are evaluated, and alternativeformulations are explored. Typically, manyscientists and engineers are involved inimproving and optimizing individualprocessing steps.

Simulation tools used at this stage canintroduce a common communicationlanguage and facilitate team interaction. A

Demetri Petrides, Alexandros Koulouris, andCharles Siletti

Demetri Petrides is president, AlexandrosKoulouris is managing director of IntelligenEurope, and Charles Siletti is director ofscheduling and planning applications at Intelligen,Inc., 2326 Morse Avenue, Scotch Plains, NJ 07076,908.654.0088, fax 908.654.3866,info@intelligen.com, www.intelligen.com.

BioPharm AUGUST 2002 3

process simulators can be used to size processequipment and supporting utilities and toestimate the required capital investment.Production transfer to existing manufacturingsites can use process simulators to evaluatevarious sites for capacity and cost and toselect the most appropriate facility. The sametype of model can be applied to theoutsourcing of manufacturing, helping toselect a contract manufacturer.

Manufacturing. In large-scale manufacturing,simulation tools are primarily used forprocess scheduling, debottlenecking, andongoing process optimization. Simulationtools capable of tracking equipment used foroverlapping batches can identify bottleneckcandidates and guide the user through thedebottlenecking effort.

Several publications are available thataddress the use of simulation for evaluatingand optimizing integrated biochemicalprocesses (1-4). We focus here on the use ofsuch tools for identifying bottlenecks,reducing cycle times, and increasing thethroughput of existing biomanufacturingfacilities.

Using Debottlenecking TheoryThe total throughput of a batch plant withina given time period is equal to the batch size(the amount of product produced per batch)multiplied by the number of batchesexecuted during that period (Equation 1).

Furthermore, because the number ofbatches is inversely proportional to the plant— or recipe — cycle time, which representsthe interval between the start of twoconsecutive batches, the plant throughputbecomes proportional to the batch sizedivided by the plant cycle time (Equation 2).

To increase plant throughput, an increasecan be made in either the batch size or thenumber of batches or both. Increasing thoseparameters, however, increases the

likelihood of running into bottlenecks(process limitations) from either theequipment or the resources. Resourcesinclude the demand for various utilities,labor, and raw materials, among others. Thebottleneck that limits the number of batches(or the plant cycle time) are known as time(or scheduling) bottlenecks. Those that limitthe batch size are known as size bottlenecks.

Equipment-related time (or scheduling)bottlenecks are identified by tracking theuses of the various equipment over time andcalculating the equipment cycle time. Figure2 shows how such information can begraphically displayed using “Equipment UseGantt Charts.” The equipment with thelongest cycle time (V-101 in Figure 2) is thetime bottleneck that determines themaximum number of batches and the plantcycle time. Auxiliary equipment, such asclean-in-place (CIP) and steam-in-place(SIP) skids, can also become timebottlenecks.

Finding the size bottleneck. Similar to plantthroughput (Equation 1), maximum batchsize (or maximum batch throughput) of acyclical processing step can be determinedas shown in Equation 3.

Batch size of a semicontinuousprocessing step (such as high-pressurehomogenization or disk-stackcentrifugation) is determined using Equation 4.

The step that yields the lowest maximumbatch size is the size bottleneck anddetermines the maximum batch size of theentire recipe. An alternative way ofidentifying size bottlenecks is by calculatingthe capacity used, the uptime, and thecombined use of the various processing steps.

Finding the scheduling bottleneck. Thescheduling bottleneck is that step that hasthe longest duration or step cycle time. Thestep duration can be estimated as shown inEquation 5.

Finding equipment use bottlenecks. Each type ofequipment is characterized by a uniquecapacity measurement (such as a reactor’svessel volume, for instance), whichdetermines the maximum load that theequipment can handle per cycle. Capacityused is defined as the fraction of anequipment’s capacity used during anoperation. For instance, if a certain vesseloperates with a maximum liquid level of1,000 L, but the vessel operates at a volumeof 500 L, the capacity used is 0.5 or 50%.For equipment that operates in continuous orsemicontinuous mode, the capacity used is

Figure 1. Benefits from the use of bioprocess simulation tools

Idea GenerationProject Screening, Strategic Planning

Process DevelopmentEvaluation of Alternatives

Common Communication Language

Facility Design and SelectionEquipment, Utility Sizing, and Design

ManufacturingOngoing Optimization, Debottlenecking

Process Scheduling, Production Planning

Development groups

Development groupsProcess engineeringCorporate environmentalManufacturing

Manufacturing

Plant throughput = Batch size × Number of batches

Plant throughput ∝ Batch sizePlant cycle time

Equation 3 Step batch size = Cycle size × Number of cyclesBatch

Equation 2

Equation 1

4 BioPharm AUGUST 2002

Process Development

based on information available in patent andtechnical literature. We used the SuperProDesigner (Intelligen) simulator in ourexample. (Readers can obtain an evaluationcopy of the software at the company website(www.intelligen.com).

Process description. Our example analyzesthe production of a therapeutic monoclonalantibody that is required in large doses. TheFigure 3 flowchart shows howapproximately 70 kg of purified product isproduced each year in 29 batches beforedebottlenecking strategies are applied. Tosimplify the flowsheet, the media, buffer,and inoculum preparation steps wereremoved from the recipe.

The volume of bioreactor broth generatedduring each batch is about 4,000 L andcontains four kg of product (the product titer is1 g/L). Total volume of the bioreactor vessel(V-101) is 6,500 L. A cycle time of 255 hours(240 hours for fermentation and 15 hours forthe turnaround) was estimated for thebioreactor. Because the bioreactor is the timebottleneck, a new batch can begin every 264hours (11 days). Figure 2 shows the equipmentuse during two consecutive batches.

Biomass and other suspended compoundsgenerated are removed using a 0.65-µmmembrane diafilter (DF-101). Productrecovery at this step is 95%. The filtration steptakes about 12 hours using a filter with amembrane area of 30 m2. Clarified solution isconcentrated 15-fold using a 50,000 MW cut-off diafilter (DF-102). Recovery yield at thisstep is 95% and takes 6.35 hours using a filterwith an 80-m2 membrane.

Concentrated product is stored in anagitated tank (V-103) with a total volume of1,200 L. The bulk of the contaminant proteinsare removed using protein A affinitychromatography (C-101). The column handlesthe material from each batch in two cycleslasting 8.8 hours each. The column has aheight of 0.25 m and a diameter of 0.9 m. Thebinding capacity of the resin is 15 mg ofproduct per mL of resin, and product recoveryis 90%. The protein A elution buffer isexchanged with phosphate buffer (P-7) usingthe same diafilter (DF-102) as in theconcentration step (P-4). The multiplerectangles on the Gantt chart at DF-102(Figure 2) show that the same equipment isused by two different unit procedures. Productrecovery at this step is 95%.

Purification continues, using a cation-exchange column (C-102), operated for two

we recommend that two approaches (thestep batch size and the combined use) beused in conjunction to identify the true batchsize bottleneck. We recommend the stepsoutlined in the “Debottlenecking Strategy”sidebar for biomanufacturing facilities

An Example of DebottleneckingA cell culture plant that produces atherapeutic monoclonal antibody (MAb) canbe used to illustrate throughput analysismethodology. MAbs are used in diagnostictests as well as for therapeutics. Worlddemand for approved MAbs is now a fewkilograms per year. However, with newtherapeutic MAbs that require dosesbetween several hundred milligrams and agram during the course of a therapy, theworld demand for MAbs is expected to riseto hundreds of kilograms per year (5).

This example illustrates how to increaseplant throughput with only modest capitalexpenditures. The flowsheet we generated is

the operating throughput divided by themaximum possible throughput for thatparticular process material or equipment.

A measure of how effectively a piece ofequipment is used is given by the equipmentuptime defined as the ratio of thatequipment’s occupancy time over the plantcycle time. For example, if the cycle time is100 hours, and a certain vessel is only usedfor 20 hours during a batch, its uptime is 0.2 or 20%. The equipment occupancy timeis the sum of the time that particularequipment takes to execute the tasks hostedin that equipment.

Combined use is the product of thecapacity used and the uptime, and combineduse is a clear indicator of how much of thetime and capacity of particular equipment isactually being used.

Using these equations as indicators, theprocessing step with the highest combineduse is most likely the batch size bottleneck.Because of time constraints in most plants,

If the goal is to increase plant throughput,changes that increase the batch size orreduce the plant cycle time can beeffective. In general, we recommend thefollowing strategy.

• Increase batch size until at least onecyclical step operates at 100% usecapacity.

• If equipment uptime is low, tryincreasing the number of cycles per batchfor that equipment. This may createopportunities for additional increases inbatch size. A side benefit of increasedbatch size is the reduced cost for qualitycontrol (QC) and quality assurance (QA),which depend on the number not the sizeof the batches.

• If a process operates at its maximumbatch size, work to reduce plant cycletime by eliminating time bottlenecks.Long process steps and equipmentsharing cause time bottlenecks.

• If bottlenecks are created by equipmentsharing install extra equipment thatreduces the sharing. The size of the newequipment should be chosen to createopportunities for batch size increases(basing the equipment size on the mostdemanding step). Rearranging the order

in which equipment is used (for sharedequipment) can create opportunities forreducing cycle time and sometimes forbatch size increases.

• If the time bottleneck is caused by astep that has a very long cycle time, newequipment should be operated in astaggered mode based on the cycle timeof the next time bottleneck.

• If the time bottleneck is caused byequipment, it can sometimes beeliminated by moving secondaryoperations from bottlenecked tononbottlenecked equipment (1). Forinstance, instead of heating material in avessel, heating can be done using anexternal heat exchanger during thecharge and transfer of material into thevessel.

• If bottleneck analysis suggests buyingnew equipment, the final purchasedecision should be based on anevaluation of overall project economiccriteria, not simply on throughputconsiderations.

Reference(1) T.M. Minnich, “Use Process Integration forPlant Modernization,” Chem. Eng., 70-76,(August 2000).

Debottlenecking Strategy

BioPharm AUGUST 2002 5

cycles lasting 8.95 hours each. The column’sheight is 0.25 m, and its diameter is 0.8 m.The binding capacity of the resin is 20 mg ofproduct per mL of resin, and the productyield is 90%. Using an agitated tank (V-103), ammonium sulfate is then added toa concentration of 2.0 M, increasing theionic strength of the solution and preparingit for hydrophobic interactionchromatography (HIC). The HIC column(C-103) handles the batch material in twocycles that each last 20.6 hours. The columnhas a height of 0.25 m and a diameter of 0.7 m, with a binding capacity of 20 mg ofproduct per mL of resin.

Sodium chloride and acetic acid areadded to inactivate any virus particles (P-11/V-104), and the solution is pushedthrough a 0.1-µm membrane diafilter (DF-103) that captures viral and othersuspended particles. As a final diafiltrationstep (P-13), reusing an existing diafilter(DF-102), the HIC and virus inactivationchemicals are replaced with PBS buffer.Final product solution is about 9.5 g/L, andthe final amount of product is 2.3 kg, whichcorresponds to an overall recovery yield of56%.

Options for increased throughput. Assumingthat the market demand for the product inour example is rising, options need to beconsidered for increasing the plant’sthroughput. Figure 4 shows the capacity,time, and combined use for allprocedure–equipment pairs as our processnow stands. The bioreactor (V-101) has thehighest combined use (62.6%) and isidentified as the first throughput bottleneckcandidate. However, the bioreactor capacityused is only 65.3% because the maximumworking volume is 6,175 L (0.95 � 6,500),whereas the broth volume is only 4,000 L,

4,0006,175

= 0.653 or 65.3%

This provides an opportunity forincreasing the batch size by 54.4%. Such amove is in agreement with thedebottlenecking strategy that recommendsan “increase in batch size until a batch sizebottleneck is reached.” When that is done,however, all three chromatography columnsbecome unable to handle the amount ofmaterial using only the two cycles that wasassumed at the beginning (the software weare using displays error messages to warnthe user about such discrepancy).

To accommodate the new batch size, thenumber of cycles per batch for all columnshas to increase from two to three.Fortunately, that increase has no negativeaffect on plant cycle time or the annualnumber of batches (which remains at 29)because of idle time in downstreamprocessing.

The bioreactor remains the time bottleneck. Thenew plant throughput is 3.5 kg of productper batch, which translates to 102.8 kg ofproduct per year. This first debottleneckingstep clearly shows that oversized equipmentoffers opportunities for increased plantthroughput without capital expenditures.Although the diafilters (DF-101, DF-102,and DF-103) are shown (Figure 4) to be at100% use capacity use, they are not thebatch size bottlenecks because their uptimesare rather low, and increasing their batchsize simply increases their uptime.

After increasing the batch size, thebioreactor (P1/V-101) is fully used both insize (capacity) and in time. At this point, thebioreactor constitutes the batch size as wellas the plant throughput bottleneck. The onlyway to increase plant throughput beyond thecurrent level is by installing extra capacityfor procedure P-1 (that is, by installing anadditional bioreactor) and by staggering theoperation of the second bioreactor so that itstarts during the middle of the firstbioreactor’s cycle time.

Making those changes are not yet feasiblebecause the DF-102, which is the new timebottleneck, has a cycle time (196 hours) thatis more than half the cycle time of the P-1(260 hours). Consequently, the startingtimes of the two bioreactors are staggered196 hours apart, and this is the new plantcycle time. Figure 5 shows an equipment usechart for the new scenario. Under these

Figure 2. Equipment use chart

Figure 4. Capacity, time, and combined usechart

Figure 3. Monoclonal antibody production flowsheet (s � stream, p � procedure, v � vessel, df � diafilter, and c � column)

6 BioPharm AUGUST 2002

Process Development

such a case, labor can be the timebottleneck.

Labor demand analyses (such as that inFigure 7) are also useful in staffing afacility. If the facility is dedicated tomanufacturing a single product, then thenumber of operators in each shift should bebased on the peak demand during that shift.In multiproduct facilities, each productionsuite can employ a dedicated number ofoperators and use floating operators duringperiods of peak demand.

Figure 8 shows the water-for-injection(WFI) demand for six consecutive batches.The blues lines represent the demand forWFI (averaged over an eight-hour period)and correspond with the y-axis on the leftside. The red lines represent the liquid level

sharing), and V-105 (a storage tank added toreplace V-103 in P-9) becomes the new timebottleneck. However, the bioreactors are againapproaching 100% uptime. The same is truefor one of the diafilters (DF-105).Furthermore, the chromatography columnsare approaching a combined use of 70%.Further throughput increase modifications areprobably impractical at this point. If additionalplant throughput is required, a new productionline should be designed and built.

Floor space. In the previous analysis, weassumed that floor space was available forinstalling a second and third bioreactor andadditional recovery equipment. If that is notthe case, then a new building may need to beconstructed, and our best throughputscenario may not necessarily be the mosteconomically attractive.

Resource bottlenecks. Another characteristicof batch processing is the variable demandfor resources (labor, utilities, and rawmaterials) as a function of time. Forinstance, Figure 7 shows the labor demand(expressed in number of operators) for sixconsecutive batches. As shown, for shortperiods, up to 17 operators are needed. Ifthat labor is unavailable, then certainoperations need to be delayed to distributethe demand for operators more evenly. In

conditions, the annual number of batchesincreases from 29 to 39 and the annualthroughput from 102.8 to 138.3 kg.

Figure 5 also shows how the twostaggered bioreactors alternate in handlingconsecutive batches. The first line (V-101)represents the first bioreactor (handling thefirst and third batches and all subsequentodd-numbered batches), and the second line(STG01>>V-101) represents the secondbioreactor (handling the second and fourthbatches and all subsequent even-numberedbatches). In other words, we use the samerecovery train to handle both bioreactors.That is possible because all recovery stepshad cycle times considerably lower than thecycle time of the bioreactor.

Removing time bottlenecks. Because theprocess now operates at its maximum batchsize, all additional debottlenecking actionsshould focus on the elimination of timebottlenecks. Figure 5 clearly shows that thediafilter DF-102 is the current timebottleneck. Addition of a new diafilter toreplace DF-102 in P-13 removes that timebottleneck and increases the number ofbatches to 44 and the annual production to156 kg. At that point, V-103 becomes thenew time bottleneck with a plant cycle timeof 171 hours. Adding an extra storage vesselto replace V-103 in P-9 increases thenumber of batches to 58 and the annualthroughput to 205.7 kg. With those changes,the bioreactors (V-101 and STG01>>V-101)again become the time and throughputbottleneck, and the new cycle time has beenlowered to 130 hours.

Plant throughput can be increased furtherwith the addition of a new bioreactor and anew diafilter (to replace DF-102 in P-7).Under those conditions, the number of batchesbecomes 84, and the annual throughput goesup to 298 kg (Figure 6). At this point all stepsnow use dedicated equipment (no equipment-

Figure 5. Equipment use chart with twobioreactors operating in staggered mode

Figure 6. Equipment use chart with threebioreactors and no equipment sharing

Equation 4 Step batch size = Continuous throughput × Plant cycle time

Equation 5 Step duration =Step cycle time

Number of units available

Equation 6 Combined use = Capacity used × Uptime

Figure 7. Labor demand as a function oftime (six consecutive batches)

Figure 8. Water-for-injection consumptionrate and storage tank levels (sixconsecutive batches)

BioPharm AUGUST 2002 7

in the WFI storage tank and correspond withthe y-axis on the right side. WFI demand is afrequent bottleneck in biopharmaceuticalmanufacturing. It is commonly used duringmultiple processing steps simultaneously toprepare fermentation media and purificationbuffers and to make equipment cleaningsolutions.

If the WFI storage tank runs dry at acertain point (because of low capacity in thestill or the tank), one or more process stepswill be delayed, and the WFI can becomethe time bottleneck. The red lines in Figure 8 (showing the liquid level in thestorage tank) stay within certain limitsbecause the WFI still, which has a capacityof 1,000 L/h, was programmed to turn onwhen the level drops below 35% and to turnoff when the level exceeds 85% of the vesselvolume of 45,000 L. During retrofit projects,such as the one described in this example,design engineers should consider new utilityprofiles to make sure adequate productionand storage capacity is available to meetnew demands. Process simulators can playan important role in retrofit projects by

enabling engineers to calculate new profilesand optimize the size of new equipment.

Tooling Up to Maximum EfficiencyUse of computer-aided tools for processdesign and simulation is increasing in thebiotech industry and affecting engineeringactivities at all stages of commercialization,from initial idea screening to processdevelopment and manufacturing.

Simulators that are capable of trackingand displaying the capacity and the time ofequipment and resource use can beindispensable in identifying and eliminatingbottlenecks. In revamping existing facilities,such tools enable management to visualizethe plant operating at maximum practicalcapacity and to evaluate processmodifications required to reach that goal. Indesigning new facilities, simulation toolscan be used to judiciously introduceovercapacity for certain steps and the abilityto expand easily for others, so that increasedfuture market demand can be readilyaccommodated.

Having access to a simulator that quicklyperforms the repetitive calculations allowsfacility engineers to focus more of their timeon creative aspects of design and to achieveoptimized solutions that would beimpossible otherwise.

References(1) D.P. Petrides, J. Calandranis, and C.L.

Cooney, “Bioprocess Optimization Via CAPDand Simulation for ProductCommercialization,” Genet. Eng. News,16(16), 24–40 (1996).

(2) F. Hwang, “Batch Pharmaceutical ProcessDesign and Simulation,” Pharma. Eng., 28-43(January–February 1997).

(3) D.P. Petrides, R. Nir, J. Calandranis, and C.L.Cooney, “Introduction to BioprocessSimulation,” Manual of IndustrialMicrobiology and Biotechnology, A.L.Demain and J.E. Davies, Eds. (ASM Press,Washington, DC, 1999), pp. 289-299.

(4) R.G. Harrison et al., Bioseparations Scienceand Engineering (Oxford University Press,New York), in press. The chapter “BioProcessDesign and Economics,” is available atwww.intelligen.com/literature.htm.

(5) S.S. Seaver, “Monoclonal Antibodies: UsingNew Techniques to Reduce DevelopmentTime,” Genet. Eng. News 17(1), 13–28 (1997).BP