batch process simulation for the pharmaceutical industry

OPTIMIZE MANUFACTURING OFPHARMACEUTICAL PRODUCTS WITHPROCESS SIMULATION AND PRODUCTIONSCHEDULING TOOLS

V. Papavasileiou, A. Koulouris, C. Siletti and D. Petrides

Abstract: This article describes how batch process simulators and scheduling tools can be usedto facilitate and expedite development and commercialization of pharmaceutical products.

Keywords: pharmaceutical manufacturing; process simulation; computer-aided process modelling;production scheduling; cost analysis; cycle time reduction; risk assessment; Monte Carlo simulation;lean manufacturing.

INTRODUCTION

The global competition in the pharmaceuticalindustry and the increasing demands by gov-ernments and citizens for affordable medicineshas focused the industrys attention on manu-facturing efciency. In this new era, improve-ments in process and product developmentapproaches and streamlining of manufacturingoperations can have a profound impact on thebottomline. Process simulation and schedulingtools can play an important role in this endea-vor. The role of such tools in the developmentand manufacturing of active pharmaceuticalingredients (APIs) has been reviewed in thepast (Petrides et al., 1996, 2002a, b; Petridesand Siletti, 2004; Thomas, 2003; Hwang,1997; Harrison et al., 2003; Tan et al., 2006).The focus of this article is on the role of suchtools in the development and manufacturingof pharmaceutical products.Common forms of pharmaceutical products

include tablets, capsules, ointments, creams,solutions in syringes and vials. Their prep-aration involves mixing of the active ingredi-ent(s) with various excipients that increasethe shelf-life of the product and facilitate thedelivery of the active ingredient. Manufactur-ing of inject able products involves lling ofsyringes or vials under aseptic conditions.During process development, process

simulation software is used to perform thefollowing tasks:

. represent the entire process on thecomputer;

. perform material and energy balances;

. estimate the size of equipment;

. calculate demand for utilities as a functionof time;

. estimate the cycle time of the process;

. perform cost analysis;

. asses the environmental impact, and so on.

The availability of a good model on thecomputer improves the understanding of theentire process by the team members andfacilitates communication. What-if and sensi-tivity analyses are greatly facilitated by suchtools. The objective of such studies is toevaluate the impact of critical parameters onvarious key performance indicators (KPIs),such as production cost, cycle times andplant throughput. If there is uncertainty for cer-tain input parameters, sensitivity analysis canbe supplemented with Monte Carlo simulationto quantify the impact of uncertainty. Costanalysis, especially capital cost estimation,facilitates decisions related to in-house manu-facturing versus outsourcing. Estimation ofthe cost-of-goods identies the expensiveprocessing steps and such information isused to guide R&D work in a judicious way.When a process is ready to move from

development to manufacturing, processsimulation facilitates technology transfer andprocess tting. A detailed computer modelprovides a thorough description of a processin a way that can be readily understood andadjusted by the recipients. Process adjust-ments are commonly required when a newprocess is moved into an existing facilitywhose equipment is not ideally sized for thenew process. The simulation model is usedto adjust batch sizes, gure out cycling ofcertain steps (for equipment that cannot

1086 Vol 85 (A7) 10861097

Correspondence to:Dr D. Petrides, Intelligen,Inc., 2326 Morse Avenue,Scotch Plains, New Jersey,USA.E-mail: [email protected]

DOI: 10.1205/cherd06240

02638762/07/$30.00 0.00

Chemical EngineeringResearch and Design

Trans IChemE,Part A, July 2007

# 2007 Institutionof Chemical Engineers

handle a batch in one cycle), estimate recipe cycle times, andso on.Production scheduling tools play an important role in man-

ufacturing (large scale as well as clinical). They are used togenerate production schedules on an on-going basis in away that does not violate constraints related to the limitedavailability of equipment, labor resources, utilities, inventoriesof materials, and so on. Production scheduling tools close thegap between ERP/MRP II tools and the plant oor (Plenertand Kirchmier, 2000). Production schedules generated byERP (Enterprise Resource Planning) and MRPII (Manufac-turing Resource Planning) tools are typically based oncoarse process representations and approximate plantcapacities and, as a result, solutions generated by thesetools may not be feasible, especially for multi-product facili-ties that operate at high capacity utilization. That oftenleads to late orders that require expediting and/or to largeinventories in order to maintain customer responsiveness.Lean manufacturing principles, such as just-in-time pro-duction, low work-in-progress (WIP), and low product inven-tories cannot be implemented without good productionscheduling tools that can accurately estimate capacity.

COMMERCIALLY AVAILABLE SIMULATION ANDSCHEDULING TOOLS

Process simulation programs, also known as processsimulators, have been in use in the chemical and petrochem-ical industries since the early 1960s. Established simulatorsfor those industries include: Aspen Plus and HYSYS fromAspen Technology, Inc. (Cambridge, MA), ChemCAD fromChemstations, Inc. (Houston, TX), and PRO/II from SimSci-Esscor, Inc. (Lake Forest, CA).The above simulators have been designed to model pri-

marily continuous processes and their transient behaviourfor process control purposes. Most pharmaceutical products,however, are produced in batch and semi-continuous mode.Such processes are best modeled with batch process simu-lators that account for time-dependency and sequencing ofevents. Batches from Batch Process Technologies, Inc.(West Lafayette, IN) was the rst simulator specic to batchprocesses. It was commercialized in the mid-1980s. All ofits operation models are dynamic and simulation alwaysinvolves integration of differential equations over a period oftime. In the mid-1990s, Aspen Technology introduced BatchPlus, a recipe-driven simulator that targeted batch pharma-ceutical processes. At around the same time, Intelligen, Inc.(Scotch Plains, NJ) introduced SuperPro Designer. The initialfocus of SuperPro was on production of APIs (synthetic andbiosynthetic) and specialty chemicals. Over the years itsscope has been extended to include modeling of processesfor the production of pharmaceutical and consumer products.Discrete-event simulators have also found applications in

the pharmaceutical industries, especially in modeling anddebottlenecking of packaging operations. Established toolsof this type include ProModel from ProModel Corporation(Orem, UT), Arena and Witness from Rockwell Automation,Inc. (Milwaukee, WI), Extend from Imagine That, Inc. (SanJose, CA), and FlexSim from FlexSim Software Products,Inc. (Orem, UT). The focus of models developed with suchtools is usually on the minute-by-minute time-dependencyof events and the animation of the process. Material bal-ances, equipment sizing, and cost analysis tasks are usually

out of the scope of such models. Some of these tools arequite customizable and third party companies occasionallyuse them as platforms to create industry-specic modules.For instance, BioPharm Services, Ltd. (Bucks, UK) havecreated a module with emphasis on biopharmaceuticalprocesses that runs on top of Extend.MS Excel from Microsoft is another common platform for

creating models for pharmaceutical processes that focus onmaterial balances, equipment sizing, and cost analysis.Some companies have even developed models in Excelthat capture the time-dependency of batch processes. Thisis typically done by writing extensive code (in the form ofmacros and subroutines) in VBA (Visual Basic for Appli-cations) that comes with Excel. K-TOPS from Alfa Laval Bio-kinetics, Inc. (Philadelphia, PA) belongs to this category.In terms of production scheduling, established tools include

Optiex from i2 Technologies, Inc. (Irving, TX), SAP APOfrom SAP AG (Walldorf, Germany), ILOG Plant PowerOpsfrom ILOG SA (Gentilly, France), Aspen SCM (formerlyAspen MIMI) from Aspen Technology, Inc. (Cambridge,MA), and so on. Their success in the pharmaceutical industry,however, has been rather limited so far. Their primary focuson discrete manufacturing (as opposed to batch chemicalmanufacturing) and their approach to scheduling from amathematical optimization viewpoint are some of the reasonsof the limited market penetration.SchedulePro from Intelligen, Inc. (Scotch Plains, NJ) is a

new nite capacity scheduling tool that focuses on schedul-ing of batch and semi-continuous chemical and relatedprocesses. It is a recipe driven tool with emphasis on gener-ation of feasible solutions that can be readily improved by theuser in an interactive manner.Examples that illustrate the benets from the use of simu-

lation and scheduling tools in the production of pharma-ceutical products follow.

MODELLING AND ANALYSIS OF A TABLETMANUFACTURING PROCESS

We will use a tablet manufacturing process as a repre-sentative example to demonstrate the use of process simu-lation and scheduling tools in the development andmanufacturing of nished pharmaceutical products. Tomodel an integrated process on the computer using Super-Pro Designer, the user starts by developing a owsheet thatrepresents the overall process. Figure 1, for instance, dis-plays part of the owsheet of a tablet manufacturing pro-cess. The owsheet is developed by putting together therequired unit procedures (see next paragraph for expla-nation), and joining them with material ow streams. Next,the user initializes the owsheet by registering the variousmaterials that are used in the process and specifyingoperating conditions and performance parameters for thevarious operations.Most pharmaceutical processes operate in batch or semi-

continuous mode. This is in contrast to petrochemical andother industries that handle large throughputs and use con-tinuous processes. In continuous operations, a piece ofequipment performs the same action all the time. In batchprocessing, on the other hand, a piece of equipment goesthrough a cycle of operations. For instance, a Slurry Prep-aration step (P-1 in V-101) includes the following operations(see Figure 2): Sanitize, Charge USP Water, Charge

Trans IChemE, Part A, Chemical Engineering Research and Design, 2007, 85(A7): 10861097

OPTIMIZE MANUFACTURING OF PHARMACEUTICAL PRODUCTS 1087

Sucrose, Charge API, Agitate, Transfer to Mill Recirculationvessel, and Flush (clean equipment). In SuperPro, the setof operations that comprise a processing step is called aunit procedure (as opposed to a unit operation). The individ-ual tasks contained in a procedure (e.g., Charge, Heat,Agitate, and so on) are called operations. A unit procedureis represented on the screen with a single equipment-lookingicon. In essence, a unit procedure is the recipe that describesthe sequence of actions required to complete a singleprocessing step. Figure 2 displays the dialogue throughwhich the recipe of a vessel unit procedure is specied. Onthe left-hand side of that dialogue, the program displays theoperations that are available in a vessel procedure; on theright-hand side, it displays the registered operations. The sig-nicance of the unit procedure is that it enables the user todescribe and model the various activities of batch processingsteps in detail.For every operation within a unit procedure the simulator

includes a mathematical model that performs material andenergy balance calculations. Based on the material balances,it performs equipment-sizing calculations. If multiple oper-ations within a unit procedure dictate different sizes for a cer-tain piece of equipment, the software reconciles the differentdemands and selects an equipment size that is appropriatefor all operations. The equipment is sized so that it is largeenough that it will not be overlled during any operation, butit is no larger than necessary (in order to minimize capital

costs). If the equipment size is specied by the user, the simu-lator checks to make sure that the vessel is not overlled. Inaddition, the tool checks to ensure that the vessel contentswill not fall below a user-specied minimum volume (e.g., aminimum stir volume) for applicable operations.In addition to material balances, equipment sizing, and cycle

time analysis, the simulator can be used to carry out cost-of-goods analysis and project economic evaluation. The sectionsthat follow provide illustrative examples of the above.Having developed a good model using a process simulator,

the user may begin experimenting on the computer withalternative process setups and operating conditions. Thishas the potential of reducing the costly and time-consuminglaboratory and pilot plant effort. Of course, the GIGO (gar-bage-in, garbage-out) principle applies to all computermodels. If critical assumptions and input data are incorrect,so will be the outcome of the simulation.When modelling an existing plant, input data required by

the model can be extracted from the data recorded by theactual process. A communication channel must thereforebe established between the modeller and the plant engin-eers. The application of some data mining technique isusually required to transform the process data to the formrequired by the model. When designing a new plant, experi-ence from similar projects can be exploited to ll-in the infor-mation gaps. In all cases, a certain level of model vericationis necessary after the model is developed. In its simplest

Figure 1. The owsheet for the pharmaceutical tablet manufacturing process.


1088 PAPAVASILEIOU et al.

form, a review of the results by an experienced engineer canplay the role of verication. Running a sensitivity analysis onkey input variables can reveal the parameters with the great-est impact on the models most important outputs. These par-ameters would then constitute the focal points in the dataacquisition effort in an attempt to estimate their values anduncertainty limits with the best possible accuracy.

Process Description

The objective of this example is to illustrate how batch pro-cess simulators can be used to model, visualize, and analysene pharmaceutical processes. It focuses on a process formanufacturing pharmaceutical tablets in an existing facilitywhose equipment sizes are known. Such tools, however,also can be used to size equipment during the design ofnew facilities.The entire owsheet of the process is shown in Figure 1. A

batch begins by charging 705 L of USP-Water into a 1500 Lmixing tank (V-101), then adding 200 kg of sucrose and400 kg of API. The suspension is agitated thoroughly for8 h. Then the suspension is transferred into another 1500 Ltank (V-102) that feeds the nano-mill (NM-101). The role ofthe nano-mill is to homogenize the suspension thoroughlyand reduce the API particles to nanometer scale. This stepis required because this specic API is insoluble in water.The suspension is pumped through the nano-mill twiceduring a period of 22 h. After the completion of the

homogenization, the material is transferred into a 2000 Ltank (V-103) where 70 kg of an excipient and 50 kg of a a-vouring agent are added along with 100 L of USP water.The material is agitated thoroughly for 5 h. Then, the materialis transferred into another 2000 L tank (V-105) that feeds thegranulator. A stabilizer solution is prepared in a 500 L tank (V-104) by dissolving 80 kg of the stabilizer into 180 L of USPwater. The stabilizer solution is combined with the homogen-ized solution in V-105.In preparation for the granulation/drying step, mannitol is

added into the bowl of the granulator (GRN-101). Then thesuspension is sprayed into the chamber of the granulatorat a rate of 120 kg h21. Almost all of the water is removed(nal water content 0.005% w/w). The granulated/driedmaterial is removed from the granulator and stored intomultiple 50 L mobile containers (MC-101). The granulatorhandles a batch of homogenized material in two cyclesbecause of its limited bowl volume that can hold up toaround 1350 L (or 600 kg) of bulk solids.The mobile tanks are moved into the tablet press room.

The tablet press (TBP-101) makes 0.5 g tablets at a rate of250 000 tablets per hour. The processing of a batch is com-pleted in approximately 9.6 h. The tablets are collected in astorage bin (DB-101).Then, a tablet coater (TB-101) is used to coat the tablets

with a material that gives them a sweet taste and a bluecolour. A batch is processed in four cycles because thecoater can handle up to around 300 kg of tablets per batch.

Figure 2. The operations associated with the rst unit procedure of Figure 1.



The coating solution is prepared ahead of time in a 150 Lmixing tank (V-106). Approximately, 10 mg of coating solutionis required per tablet. A coating cycle takes around 6 h. Warmair is pumped through the drum of the coater during the coat-ing process to vaporize the water of the coating solution. Thecoated tablets are stored in drums (not shown in the ow-sheet) and taken to the packaging area.All the unit procedures up to the granulator (P-7/GRN-101)

require a sanitization operation prior to the main processingactivities and a CIP (clean in place) operation after themain processing activities.Table 1 provides information on raw material requirements

for the entire process. A batch consists of approximately 2.4million product tablets that have a total mass of around1224 kg. The API is approximately 33% of the nal tabletmass. Notice the large amount of Hot-USP-Water, which isconsumed for equipment cleaning.

Process Scheduling and Resource Tracking

Figure 3 displays the Equipment Occupancy chart for fourconsecutive batches (each colour represents a differentbatch). The recipe scheduling summary dialogue is shownon the top right hand corner. The recipe batch time is approxi-mately 102 h. This is the total time between the start of therst step of a batch and the end of the last step of thatbatch. However, since most of the equipment items areutilized for shorter periods within a batch, a new batch isinitiated every 30 h, which is known as the recipe cycletime. The minimum possible recipe cycle time is 29.6 h.Multiple bars of the same colour on the same line representreuse (sharing) of equipment by multiple procedures or oper-ations. The single CIP skid (top line in Figure 3) is the onlyshared equipment in this process. White space between pro-cedure bars represents idle time. White space within a pro-cedure bar represents waiting time. For instance, the whitespaces in the procedure bars of V-103 and V-104 representwaiting for cleaning because of the constraints imposed bythe single CIP skid. This type of charts is an invaluable toolfor visualizing cycle times and scheduling bottlenecks.Figure 4 displays the Operations Gantt chart which

provides more detailed scheduling information. Notice,for instance, the duration of the TRANSFER-OUT-1 operationin P-6 (V-105) that feeds material to the granulator(P-7/GRN-101). The granulator handles a batch in twocycles due to its limited bowl volume. The duration ofTRANSFER-OUT-1 in P-6 is from the start of the rst granu-lation operation in P-7 to the end of the second granulationoperation in P-7. In reality there are two shorter transfer-out

Table 1. Material requirements.

Material kg/batch

USP Water 1 049Sucrose 200API-1 400Stabilizer 80Excipient-1 70Mannitol 400Flavoring 50OpaDry 24Hot-USP-Water 47 454Total 49 727

Figure 3. Equipment occupancy chart for four consecutive batches. This gure is available in colour online via www.icheme.org/cherd



operations (synchronized to the two granulation operations inP-7) but for simplicity SuperPro represents that with a singletransfer-out operation that has the correct total duration.Process scheduling in the context of a simulator is fully pro-

cess driven and the impact of changes can be analysed in amatter of seconds. For instance, the impact of an increase inbatch size (that affects the duration of charge, milling, granu-lation, and other scale-dependent operations) on the recipecycle time and the number of batches can be evaluatedinstantly. Due to the many interacting factors involved witheven a relatively simple process, simulation tools that allowusers to describe their processes in detail, and to quickly per-form what-if analyses, can be extremely useful.Another characteristic of batch processing is the variable

demand for resources (e.g., labour, utilities and rawmaterials) as a function of time. For instance, Figure 5 dis-plays the demand for puried water (USP water) for eightconsecutive batches. This demand includes USP water con-sumed for the CIP operations as well as USP water utilizedfor preparing the product mixture. The red lines representthe instantaneous demand whereas the green line representsthe cumulative demand and corresponds to the y-axis on theright-hand side. The blue line corresponds to daily demand(the averaging period can be adjusted by the user). Highpurity water is a common potential bottleneck in pharma-ceutical processes. It is frequently used by multiple proces-sing steps simultaneously, in activities such as solutionpreparation and equipment cleaning. If not enough instan-taneous (or cumulative) capacity is available, one or moreprocess steps may be delayed, possibly with severe

consequences. During design of new facilities or retrot ofexisting ones, the information provided by such charts isused for sizing utility systems.In addition to instantaneous demand of resources, the

simulator provides the means to track the volumetric utiliz-ation of all vessels throughout the batch cycle. This allowsthe user to track maximum working volumes over time, andensure that the minimum stir volume is always met at any rel-evant point in a process. The volume content of vessels isalso used in sizing new vessels. For existing vessels, it deter-mines their capacity utilization by all procedures executed inthem. This, in turn, identies the equipment-procedure pairthat constitutes the size bottleneck and determines the maxi-mum possible batch size that the plant can undertake. Anyeffort to exploit the economy of scale through larger batchesshould rst concentrate in removing the size bottleneck.Increasing the number of cycles for the offending procedure(i.e., split the batch in equal portions and process themseparately through the bottleneck equipment), reshufingequipment between similar-type procedures based on theircapacity needs or introducing bigger equipment are someof the ways by which a size bottleneck can be removed.

Cost Analysis

Cost analysis and project economic evaluation are import-ant for a number of reasons. For a new product, if the com-pany lacks a suitable manufacturing facility with availablecapacity, it must decide whether to build a new plant or out-source the production. Building a new plant is a major capital

Figure 4. The operations Gantt chart. This gure is available in colour online via www.icheme.org/cherd



expenditure and a lengthy process. To make a decision, man-agement must have information on capital investmentrequired and time to complete the facility. When productionis outsourced, a cost-of-goods analysis serves as a basisfor negotiation with contract manufacturers. A sufcientlydetailed computer model can be used as the basis for the dis-cussion and negotiation of the terms. Contract manufacturersusually base their estimates on requirements of equipmentutilization and labor per batch, which is information that isprovided by a good model. SuperPro performs thoroughcost analysis and project economic evaluation calculations.It estimates capital as well as operating cost. The cost ofequipment is estimated using built-in cost correlations thatare based on data derived from a number of vendors and lit-erature sources. The xed capital investment is estimatedbased on equipment cost and using various multipliers,some of which are equipment specic (e.g., installationcost) while others are process specic (e.g., cost of piping,buildings, and so on). The approach is described in detailin the literature (Harrison et al., 2003). The rest of this sectionprovides a summary of the cost analysis results for thisexample process.Table 2 shows the key economic evaluation results for this

project. The xed capital investment for a manufacturing facil-ity of this size is around $145 million. The annual operatingcost (assuming the facility is dedicated to a single product)is around $87 million, resulting in a unit production cost of

$0.14/tablet. Assuming a product selling price of $0.25/tablet, the facility generates annual revenues of $160 millionand has an attractive return on investment.Table 3 provides a breakdown of the manufacturing cost,

including (two left columns) and excluding (two right col-umns) the cost of API. A purchasing price of $500/kgwas assumed for the API, resulting in an annual API costof $52.2 million. When the cost of API is considered inthe cost-of-goods, then, the cost of raw materials becomesthe dominant cost (63% of total). If the cost of API isignored, then, the facility overhead becomes the dominantitem of the manufacturing cost (75% of total). In the lattercase, the unit production cost drops to $0.05/tablet. That

Figure 5. USP-Water demand in eight consecutive batches. This gure is available in colour online via www.icheme.org/cherd

Table 2. Key economic evaluation results.

Capital investment 144 939 000 $Operating cost 86 898 000 $/yearProduction rate 638 960 000 Tablets/yearUnit production cost 0.14 $/tabletSelling price 0.25 $/tabletTotal revenues 159 740 000 $/yearGross margin 45.60 %Return on investment 39.28 %Payback time 2.55 YearsIRR (after taxes) 32.58 %NPV (at 7.0% interest) 280 709 000 $



is representative of the incremental cost of contractmanufacturers.

Cycle Time Reduction

In a batch manufacturing facility, the annual throughput isequal to the batch size times the number of batches thatcan be processed per year. Consequently, increasing thebatch size or the number of batches per year increases theannual plant throughput.Process simulation tools enable users to readily exper-

iment with options that have the potential of increasing thebatch size and/or reducing the cycle time. The base caseprocess already operates at its maximum batch size

(imposed by the granulator). Consequently, the only optionfor throughput increase is by reducing the cycle time of thetime bottleneck equipment, which is the bin (DB-101) thatsupplies material to the tablet coater (TB-101). The avail-ability of a second bin of the same size (e.g., DB-101b) thathandles alternating batches will eliminate that bottleneckand shorten the recipe cycle time to 27.4 h (the cycle timeof the base case is 30 h). The addition of a new bin shiftsthe time bottleneck to V-105, the tank that feeds the granula-tor. Addition of a new tank of the same size (e.g., V-105b)eliminates that bottleneck, reduces the recipe cycle time to26.5 h, and shifts the bottleneck to V-102 (the tank thatfeeds the nano-mill). Finally, addition of a new nano-mill feed-ing tank (V-102b) reduces the cycle time to 26 h. Figure 6shows the equipment occupancy chart after the addition ofthe new equipment (DB-101b, V-105b, and V-102b). Thenew cycle time is 13.3% shorter than the original. Is it worthinstalling two new blending tanks and a bin for a 13.3%cycle time reduction? A denite answer to the question canbe provided by performing a cost-and-benets analysisusing the simulator. Additional information on cycle timereduction and debottlenecking approaches and method-ologies can be found in the literature (Petrides et al., 2002b).

Uncertainty and Variability Analysis

Process simulation tools typically used for batch processdesign, debottlenecking, and cost estimation employ

Table 3. Breakdown of the annual manufacturing cost.

Including cost of API Excluding cost of API

Cost item $/year % $/year %

Raw Materials 54 805 000 63.07 2 605 000 7.51Labor 4 653 000 5.35 4 653 000 13.41Facility-Overhead 26 048 000 29.98 26 048 000 75.07Laboratory/QC/QA 1 163 000 1.34 1 163 000 3.35Consumables 67 000 0.08 67 000 0.19Utilities 162 000 0.19 162 000 0.47Total 86 898 000 100.00 34 698 000 100.00

Figure 6. The equipment occupancy chart after the addition of the new equipment. This gure is available in colour online via www.icheme.org/cherd



deterministic models. They model the average or expectedsituation commonly referred to as the base case ormost likelyscenario. Modeling many cases can help determine therange of performance with respect to key process par-ameters, however, such an approach does not account forthe relative likelihood of the various cases. Monte Carlo simu-lation is a practical means of quantifying the risk associatedwith uncertainty in process parameters. In a Monte Carlosimulation, uncertain input variables are represented withprobability distributions. A simulation calculates numerousscenarios of a model by repeatedly picking values from auser dened probability distribution for the uncertain vari-ables and using those values for the model to calculate andanalyse the outputs in a statistical way in order to quantifyrisk. The outcome of this analysis is the estimation of thecondence by which desired values of key performanceindicators can be achieved. Inversely, the analysis can helpidentify the input parameters with the greatest effect on thebottomline and the input value ranges that minimize outputuncertainty.In batch, and especially pharmaceutical, processing uncer-

tainty can emerge in operation or market-related parameters.Process times, equipment sizes, material purchasing andproduct selling prices are common uncertain variables. Thepressure in the pharmaceutical industry to make new com-pounds available to patients as soon as possible meansthat process design has to be performed in early phaseswhere, however, the uncertainty is greater. Performing a sto-chastic analysis early on in the design phase increases themodels robustness and minimizes the risk of encounteringunpleasant surprises later on.For models developed in SuperPro, Monte Carlo simu-

lation can be performed by combining SuperPro with CrystalBall from Decissioneering, Inc. (Denver, Colorado). CrystalBall is an Excel add-in application that facilitates MonteCarlo simulation. It enables the user to designate the uncer-tain input variables, specify their probability distributions andselect the output (decision) variables whose values arerecorded and analysed during the simulation. For each simu-lation trial (scenario) Crystal Ball generates random valuesfor the uncertain input variables selected in frequency dic-tated by their probability distributions using the Monte Carlomethod. Crystal Ball also calculates the uncertainty involvedin the outputs in terms of their statistical properties, mean,median, mode, variance, standard deviation and frequencydistribution.As explained in the previous section, the storage units that

supply material to the tablet coater, the granulator, and thenano-mill have the longest cycle times and therefore arethe most likely scheduling bottlenecks. Since the processoperates with a cycle time of 30 h which is very close tothe minimum of 29.6 h, variability in the tablet coating, gran-ulation, and nano-milling operations that determine thecycle times of their feeding units may have an impact onthe recipe cycle time. Likely sources of variability includerandom power outages and equipment failures, availabilityof operators, differences in skills of operators that affectsetup and operation of equipment, and so on. For illustrationpurpose, the Weibull distributions of Figure 7(ac) wereassumed for the process times of tablet coating, granulation,and nano-milling, respectively. Their base case values are6 h, 7.6 h and 21.8 h, respectively. Please note that thecoater has the longest cycle time among the above three

because it handles a batch in four cycles and each cyclelasts 6 h. If this type of analysis is done for an existing facility,historical data should be used to derive the probabilitydistributions.The decision (output) variable considered in this example

is the recipe cycle time that determines the throughput ofthe facility. Figure 8 displays the results of the Monte Carlosimulation. The analysis reveals that the process can operatewith a cycle time of less than 30 h with a certainty of 80%(blue area of Figure 8). If campaign production plans arebased on the base case cycle time, the above analysisreveals that production can be completed on time with a cer-tainty of only 80%. Such ndings can assist the managementof a company in the decision making process. Additionalinformation on Monte Carlo simulation and risk assessmentcan be found in the literature (Achileos et al., 2006).

PRODUCTION PLANNING AND SCHEDULING

Pharmaceutical manufacturing facilities are typicallyequipped with multiple production lines that share utilitiesand labour resources. They may also share auxiliaryequipment, such as CIP skids, transfer panels, and deliverylines, and occasionally main equipment. They operate24/7 or with other shift patterns. Production is typically

Figure 7. Assumed probability distributions for the process times oftablet coating (a), granulation (b) and nano-milling (c).



campaigned. Considerable changeover time is often requiredbetween campaigns of different products.Scheduling tools employed in the pharmaceutical industry

must be able to quickly generate feasible solutions thatrespect all the major constraints. Such tools must beequipped with intuitive interfaces that enable the schedulerto visualize and easily modify the schedule. For instance, ifa machine goes down or an operation is delayed, the schedu-ler must be able to quickly update the state of the system andgenerate a new feasible schedule.Being able to perform what-if studies for capacity analysis

exercises is another desirable feature. This is typically doneby evaluating hypothetical production scenarios over aperiod of months to years. Such analyses can provide justi-cations for facility expansions and/or outsourcing of pro-duction (if the current facility cannot meet the expectedfuture demand).Results of scheduling tools are communicated to stake-

holders through various charts and reports that provideinformation on tasks that must be executed during a certaintime period. Figure 9 displays a production schedule in theform of a Gantt chart generated by SchedulePro. It corre-sponds to a tablet manufacturing facility equipped withtwo production lines (A and B). The two production linesutilize dedicated main equipment but share two cleaning-in-place (CIP) skids represented by the top two lines ofthe chart. The CIP skids can be used to clean equipmentin either of the two lines. Line A includes a tablet coater,but not line B. Line A operates seven days a week whereasline B operates ve days a week. The grey columns in lineB represent 48-h weekend breaks. A 24-h changeover timeis assumed between campaigns of different products toaccount for equipment adjustments and facility sanitization.For instance, the white rectangles between the rst(blue colour rectangles) and second (magenta colour rec-tangles) campaigns of line A represent such a 24-h change-over time.A production schedule is readily updated through the

chart. Updates are required when a piece of equipment

goes down or when the completion of a task is delayed.For the latter case, the user simply species the new com-pletion time of the delayed operation and adjusts the sche-dule to eliminate conicts. The adjustment can be doneautomatically by the tool or manually by the user by drag-ging and dropping activities in an interactive manner. Inessence the chart becomes an intelligent electronic LegoBoard whose activities are linked and conicts are readilyidentied and eliminated. Lego boards and other mechan-isms of manual scheduling are still prevalent in manypharmaceutical manufacturing facilities. Interactive sche-duling tools that resemble the look of Lego boards areusually adopted without much resistance by manufacturingpersonnel.A production schedule is often constrained not by its main

equipment, but instead by the availability of labor and otherresources. For instance, Figure 10 displays the labourdemand for line A of the schedule of Figure 9. The bluelines represent the total instantaneous demand for labour.The red line represents the maximum number of oper-

ators (10 in this case) in that production line. For shortperiods of time there is a need for a total of eleven oper-ators. Scheduling tools enable manufacturing personnel toreadily visualize and resolve such conicts. The resolutionof such conicts is accomplished either by the schedulingalgorithm or the user. It typically involves the delay ofsome operations that contribute to the peaks. If the facilityemploys few oating operators for lines A and B, such con-icts also can be resolved by utilizing the oating operatorsduring peak demand periods.Constraints imposed by inventories of raw materials, inter-

mediates, and nal products and waste materials are handledin a similar manner. The tool calculates the level of materialsand schedules accordingly (to avoid conicts) or simplywarns the user of conicts due to inventories and lets him/her take corrective action.In summary, scheduling tools enable manufacturing

personnel to maintain a dynamic model of the entirefacility that evolves with time and facilitates generation of

Figure 8. Calculated probability distribution for the cycle time of the entire process. This gure is available in colour online via www.icheme.org/cherd



Figure 10. Line-A labour demand as a function of time. This gure is available online via www.icheme.org/cherd

Figure 9. Production schedule in Gantt chart format. This gure is available online via www.icheme.org/cherd



production schedules that are feasible and easilymodiable. The end result is increased productivity,improved customer service, and reduced manufacturingcost.

SUMMARY

Process simulation and production scheduling toolscan play an important role throughout the life-cycle ofproduct development and commercialization. In processdevelopment, process simulation tools are becomingincreasingly useful as a means to analyse, communicateand document process changes. During the transitionfrom development to manufacturing, they facilitate technol-ogy transfer and process tting. Production scheduling toolsplay a valuable role in manufacturing. They are used togenerate production schedules based on the accurate esti-mation of plant capacity, thus minimizing late orders andreducing inventories. Such tools also facilitate capacityanalysis and debottlenecking tasks.The pharmaceutical industry has only recently begun

making signicant use of process simulation and schedulingtools. Increasingly, universities are incorporating the use ofsuch tools in their curricula. In the future, we can expect tosee increased use of these technologies and tighter inte-gration with other enabling IT technologies, such as supplychain tools, manufacturing execution systems (MES), batchprocess control systems, process analytics tools (PAT), andso on. The result will be more robust processes and efcientmanufacturing leading to more affordable medicines.

REFERENCES

Achilleos, E.C., Calandranis, J.C., and Petrides, D.P., 2006,Quantifying the impact of uncertain parameters in the batch manu-facturing of active pharmaceutical ingredients, PharmaceuticalEngineering, 3440.

Harrison, R.G., Todd, P., Rudge, S.R. and Petrides, D.P., 2003,Bioseparations Science and Engineering (Oxford UniversityPress).

Hwang, F., 1997, Batch pharmaceutical process design and simu-lation, Pharmaceutical Engineering, 2843.

Petrides, D.P., Koulouris, A. and Lagonikos, P.T., 2002a, The role ofprocess simulation in pharmaceutical process development andproduct commercialization, Pharmaceutical Engineering, 22(1): 1.

Petrides, D., Koulouris, A. and Siletti, C., 2002b, Throughput analysisand debottlenecking of biomanufacturing facilities, a job for pro-cess simulators, BioPharm.

Petrides, D.P. and Siletti, C.A., 2004, The role of process simulationand scheduling tools in the development and manufacturing ofbiopharmaceutical, Proceedings of the 2004 Winter SimulationConference, Ingalls, R.G., Rossetti, M.D., Smith, J.S. and Peters,B.A. (eds). 20462051.

Petrides, D.P., Calandranis, J. and Cooney, C.L., 1996. Bioprocessoptimization via CAPD and simulation for product commercializa-tion, Genetic Engineering News, 16(16): 2440.

Plenert, G. and Kirchmier, B., 2000, Finite Capacity Scheduling Management, Selection, and Implementation (John Wiley & Sons).

Tan, J., Foo, D.C.Y., Kumaresan, S. and Aziz, R.A., 2006, Debottle-necking of a batch pharmaceutical cream production, Pharma-ceutical Engineering, 7282.

Thomas, C.J., 2003, A design approach to biotech process simu-lations, BioProcess International, 29.

The manuscript was received 11 December 2006 and accepted forpublication after revision 6 February 2007.



batch process simulation for the pharmaceutical industry

Documents

process development

batch process simulators

production cost

impact of uncertainty

capital cost estimation

estimation ofthe cost

able products

profound impact