ARTICLE IN PRESS

0925-5273/$ - see

doi:10.1016/j.ijp

�Correspondifax: +1412 624

E-mail addre

Int. J. Production Economics 107 (2007) 223–236

www.elsevier.com/locate/ijpe

Analyzing the benefits of lean manufacturing and value streammapping via simulation: A process sector case study

Fawaz A. Abdulmaleka, Jayant Rajgopalb,�

aIndustrial and Management Systems Engineering Department, Kuwait University, KuwaitbDepartment of Industrial Engineering, University of Pittsburgh, Pittsburgh, PA 15261, USA

Received 1 November 2005; accepted 1 September 2006

Available online 28 November 2006

Abstract

The ‘‘lean’’ approach has been applied more frequently in discrete manufacturing than in the continuous/process sector,

mainly because of several perceived barriers in the latter environment that have caused managers to be reluctant to make

the required commitment. We describe a case where lean principles were adapted for the process sector for application at a

large integrated steel mill. Value stream mapping was the main tool used to identify the opportunities for various lean

techniques. We also describe a simulation model that was developed to contrast the ‘‘before’’ and ‘‘after’’ scenarios in

detail, in order to illustrate to managers potential benefits such as reduced production lead-time and lower work-in-process

inventory.

r 2006 Elsevier B.V. All rights reserved.

Keywords: Lean manufacturing; Value stream mapping; Simulation; Process industries; Steel

1. Introduction

Lean manufacturing is one of the initiatives thatmany major businesses in the United States havebeen trying to adopt in order to remain competitivein an increasingly global market. The focus of theapproach is on cost reduction by eliminating non-value added activities. Originating from the ToyotaProduction System, many of the tools and techni-ques of lean manufacturing (e.g., just-in-time (JIT),cellular manufacturing, total productive mainte-nance, single-minute exchange of dies, productionsmoothing) have been widely used in discrete

front matter r 2006 Elsevier B.V. All rights reserved

e.2006.09.009

ng author. Tel.: +1412 624 9840;

9831.

ss: rajgopal@pitt.edu (J. Rajgopal).

manufacturing. Applications have spanned manysectors including automotive, electronics, whitegoods, and consumer products manufacturing.

On the other hand, applications of lean manufac-turing in the continuous process sector have been farfewer (Abdullah and Rajgopal, 2003). It has some-times been argued that in part, this is because suchindustries are inherently more efficient and have arelatively less urgent need for major improvementactivities. Managers have also been hesitant to adoptlean manufacturing tools and techniques to thecontinuous sector because of other characteristicsthat are typical in this sector. These include large,inflexible machines, long setup times, and the generaldifficulty in producing in small batches.

While some lean manufacturing tools mightindeed be difficult to adapt to the continuous sector,

.

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236224

this does not mean that the approach is completelyinapplicable; for example, Ahmad et al. (2005),Melton (2005), Radnor (2000), Cook and Rogowski(1996), and Billesbach (1994). Abdullah et al. (2002)and Abdelmalek et al. (2006) examine aspects ofcontinuous production that are amenable to leantechniques and present a classification scheme toguide lean implementation in this sector. Theobjective of this paper is to use a case-basedapproach to demonstrate how lean manufacturingtools when used appropriately, can help the processindustry eliminate waste, maintain better inventorycontrol, improve product quality, and obtain betteroverall financial and operational control. A largeintegrated steel mill is used to illustrate theapproach followed. Since some of the informationis confidential, the company is referred to as ABsteel (or ABS) throughout this paper. In ourapproach, value stream mapping (VSM) is firstused to map the current operating state for ABS.This map is used to identify sources of waste and toidentify lean tools for reducing the waste. A futurestate map is then developed for the system with leantools applied to it. Since the implementation of therecommendations is likely to be both expensive andtime-consuming, we develop a simulation model forthe managers at ABS in order to quantify thebenefits gained from using lean tools and techni-ques.

2. Background

We begin by providing a brief overview of theprinciples used in this work, followed by somebackground information on the company where thework was conducted.

2.1. Overview of lean manufacturing and its tools

After World War II Japanese manufacturers werefaced with vast shortages of material, financial, andhuman resources. These conditions resulted inthe birth of the ‘‘lean’’ manufacturing concept(Womack et al., 1990). Kiichiro Toyoda, thepresident of Toyota Motor Company at the time,recognized that American automakers of that erawere out-producing their Japanese counterparts bya factor of about ten. Early Japanese industrialleaders such as Toyoda, Shigeo Shingo, and TaiichiOhno responded by devising a new, disciplined,process-oriented system, which is known today asthe ‘‘Toyota Production System,’’ or ‘‘Lean Man-

ufacturing.’’ The system focused on pinpointingthe major sources of waste, and then using toolssuch as JIT, production smoothing, setup reductionand others to eliminate the waste. A very briefdescription of the most common lean tools isgiven below (Monden, 1998; Feld, 2000; Nahmias,2001); the interested reader is referred to one of themany books on lean manufacturing for moredetails:

�

Cellular manufacturing: Organizes the entireprocess for a particular product or similarproducts into a group (or ‘‘cell’’), including allthe necessary machines, equipment and opera-tors. Resources within cells are arranged to easilyfacilitate all operations. � Just-in-time (JIT): A system where a customerinitiates demand, and the demand is thentransmitted backward from the final assemblyall the way to raw material, thus ‘‘pulling’’ allrequirements just when they are required. � Kanbans: A signaling system for implementingJIT production. � Total preventive maintenance (TPM): Workerscarry out regular equipment maintenance todetect any anomalies. The focus is changed fromfixing breakdowns to preventing them. Sinceoperators are the closest to the machines, theyare included in maintenance and monitoringactivities in order to prevent and provide warningof malfunctions. � Setup time reduction: Continuously try to reducethe setup time on a machine. � Total quality management (TQM): A system ofcontinuous improvement employing participativemanagement that is centered on the needs ofcustomers. Key components are employee in-volvement and training, problem-solving teams,statistical methods, long-term goals, and recogni-tion that inefficiencies are produced by thesystem, not people. � 5S: Focuses on effective work place organizationand standardized work procedures.

2.2. Overview of VSM

A value stream is a collection of all actions (value-added as well as non-value-added) that are requiredto bring a product (or a group of products that usethe same resources) through the main flows, starting

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 225

with raw material and ending with the customer(Rother and Shook, 1999). These actions considerthe flow of both information and materials withinthe overall supply chain. The ultimate goal of VSMis to identify all types of waste in the valuestream and to take steps to try and eliminate these(Rother and Shook, 1999). While researchershave developed a number of tools to optimizeindividual operations within a supply chain, most ofthese tools fall short in linking and visualizing thenature of the material and information flowthroughout the company’s entire supply chain.Taking the value stream viewpoint means workingon the big picture and not individual processes.VSM creates a common basis for the productionprocess, thus facilitating more thoughtful decisionsto improve the value stream (McDonald et al.,2002).

VSM is a pencil and paper tool, which is createdusing a predefined set of standardized icons (thereader is referred to Rother and Shook, 1999 fordetails). The first step is to choose a particularproduct or product family as the target forimprovement. The next step is to draw a currentstate map that is essentially a snapshot capturinghow things are currently being done. This isaccomplished while walking along the actualprocess, and provides one with a basis for analyzingthe system and identifying its weaknesses. The thirdstep in VSM is to create the future state map, whichis a picture of how the system should look after theinefficiencies in it have been removed. Creating afuture state map is done by answering a set ofquestions on issues related to efficiency, and ontechnical implementation related to the use of leantools. This map then becomes the basis for makingthe necessary changes to the system.

2.3. Simulation in support of VSM

For companies that have long relied on tradi-tional approaches to their manufacturing systems, itis often difficult to gain from management thecommitment required to implement lean manufac-turing. Doing so is hard because of differences in anumber of aspects including raw material procure-ment, inventory management, employee manage-ment, and production control. For traditionalmanufacturers, the reluctance to implement manylean ideas arises because their distinctive require-ments often make it hard to predict the magnitudeof the gains that can be achieved by implementing

these. As a result, management decisions onimplementing lean manufacturing often come downto their ‘‘belief’’ in lean manufacturing, reportedresults of others who have implemented leantechniques, and heuristic rules of thumb on theexpected payback. For many managers this isinsufficient justification, and lacks the quantifiableevidence needed to convince them to adopt lean(Detty and Yingling, 2000). This raises the questionof how we can make lean and VSM more viable.

While in some situations the future state map canbe evaluated with relatively modest effort, it is notas easy to do so in many others. For example,predicting inventory levels throughout the produc-tion process is usually impossible with only a futurestate map, because with a static model one cannotobserve how inventory levels will vary for differentscenarios (McDonald et al., 2002). In general, weneed a complementary tool with VSM that canquantify the gains during the early planning andassessment stages. An obvious tool is simulation,which is capable of generating resource require-ments and performance statistics whilst remainingflexible to specific organizational details. It can beused to handle uncertainty and create dynamicviews of inventory levels, lead-times, and machineutilization for different future state maps. Thisenables the quantification of payback derived fromusing the principles of lean manufacturing, andthe impact of the latter on the total system. Theinformation provided by the simulation can enablemanagement to compare the expected performanceof the lean system relative to that of the existingsystem it is designed to replace (Detty and Yingling,2000), and assuming that this is significantly super-ior, it provides a convincing basis for the adoptionof lean.

2.4. Company and process background

ABS produces several grades of steel that are usedprimarily in appliance manufacturing. The focus ofthis VSM is on one product family: annealedproducts, of which there are three types produced;open coil annealed, hydrogen batch annealed, andcontinuous annealed. Average customer demandwas estimated as 76,500 tons per month, and thedistribution by product is as follows:

�

8500 tons per month of open coil (OCA), � 10,000 tons per month of continuous (CA), � 58,000 tons per month of hydrogen batch (HBA).

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236226

The processes for this product family start with ablast furnace where on a daily basis raw material

including skips of iron ore, coke, and limestone arecharged at the top of the furnace. The melted rawmaterial is then poured into sub-ladles (essentially,large bins for holding liquid iron) from the tap holeat the bottom of the furnace. The liquid iron travelsin the sub-ladle to the basic oxygen process (BOP)where scrap is added and oxygen is blown in to burnoff excess carbon and obtain the initial form ofliquid steel. Depending on the grade of the final steelto be produced this initial liquid steel can go eitherto a ladle metallurgical facility (LMF) or a Degasserto further refine and remove impurities from theliquid steel. The refined liquid steel then goes to adual-strand continuous caster where steel slabs arecast in accordance with specific customer widths.The hot slabs are then shipped on railroad and rackcars from the continuous caster process to thefinishing mill facility for further refining processes,which include the hot strip mill (HSM), pickling,cold reduction (CR), annealing (OCA, HBA orCA), temper mill (TM), and finally, shipping.

At ABS the business planning department re-ceives demands from two types of customers: repeatand spot business (open market). The repeatdemand is received on a weekly basis, where majorABS customers call or send through EDI theirrequirements for the weeks ahead. Since these arecommitted customers the quantity and the orderdelivery time are more or less fixed. On the otherhand, spot customers generate daily schedules.There are currently two separate scheduling groups:one is for the hot end liquid steel, which usuallyincludes the blast furnace and caster, and the secondis for the finishing mill, which handles the productfrom the HSM through shipping. When an orderarrives, business planning enters it into the planningsystem, estimates the date by which they think theycan complete it, and rough-schedule orders on theproduction units on a weekly basis. Next, they affixa routing on the order and assign a ‘‘plan week’’ toit. This schedule on the operating side becomes thebasis to monitor day-by-day and week-by-weekincrements against how closely they are in accor-dance with the schedule. The schedules can then beupdated further on an as-needed basis to daily oreven bi-daily schedules. ABS uses three types oftransportation modes: truck, rail, and barge. Theshipments go to different customers on a daily orweekly basis. The plant works on a continuous basisfor 24 h a day all year long except for major

shutdowns and runs a three-shift operation in allproduction departments except for continuousannealing, which runs two shifts. Each shift is 8 hlong.

3. VSM: current state map

All data for the current state map were collectedaccording to the approach recommended by Rotherand Shook (1999). Data collection for the materialflow started at the shipping department, and workedbackward all the way to the blast furnace process,gathering snapshot data such as inventory levelsbefore each process, process cycle times (CTs),number of workers, and changeover (CO) times.Fig. 1 shows the current state map that wasconstructed; the small boxes in the map representthe process and the number inside the box is thenumber of workers at each process. Also, eachprocess has a data box below, which contains theprocess CT, machine reliability (MR), the numberof shifts, and the CO time. It should be noted thatthis data was collected whilst walking the shop floorand talking to the foreman and operators at eachworkstation. The processing and set-up times are allbased on the average of historical data.

Note that there are two inventory triangles aheadof some processes, one for annealed products andone for all other products. This simply indicates thatother products could be scheduled to use the processin addition to the annealed products consideredherein, so that the total inventory is actually higherthan what is shown. After collecting all theinformation and material flows, they are connectedas indicated by arrows in the map, representing howeach workstation receives its schedule from businessplanning.

The timeline at the bottom of the current statemap in Fig. 1 has two components. The firstcomponent is the production waiting time (in days),which is obtained by summing the lead-timenumbers from each inventory triangle before eachprocess. The time for one inventory triangle iscalculated by dividing the inventory quantity intothe daily customer requirements. For example, thelead-time for the inventory triangle ahead ofpickling is 17.65 days; this is calculated by dividing45,000 tons (the total inventory ahead of thepickling) by 2550 (the daily average demand ratefor the annealed product). The total observed valuefor the waiting time is around 46 days. Other thanabout three days that are required for the coils to

ARTICLE IN PRESS

Fig. 1. Current state map.

F.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 227

cool down after processing at the HSM, the rest ofthis time is non-value added time. Note that we donot consider the amount of raw material at thebeginning of the production, since ABS owns themines and raw material sources, so that and rawmaterial is not an issue for them.

The second element of the timeline is theprocessing (or value-added) time, which is abouttwo days. This time is calculated by adding theprocessing time for each process in the value stream.The CT for each process is the average CT, whichwas determined by using actual data from thecompany. Thus the total lead time is around 48days. If we are conservative and include theapproximately three days required for the coils tocool down after processing at the HSM we get atotal of about five days (429,030 s) of value-addedtime; this works out to slightly over 10% of the totalproduction lead time.

4. VSM: future state map

The process of defining and describing the futurestate map starts while developing the current state

map, where target areas for improvement start toshow up. Looking at the current state map for ABSseveral things stand out: (a) large inventories, (b) thedifference between the total production lead-time(around 51 days) and the value added time (5 days),which is under 10% of the total, and (c) eachprocess producing to its own schedule. Inventoryand lead time may be viewed as two related issuessince the more the inventory, the longer any itemmust wait for its turn and thus, the longer the leadtime. In creating the ideal future state map we try toidentify lean manufacturing tools to drive both ofthese down, while looking at the schedule across theentire value stream. We follow a systematicprocedure where we try to answer a series ofstructured questions; this allows us to come up withan ideal future state map that will help in eliminat-ing or at least reducing different types of waste inthe current manufacturing system.

Question 1. What is the takt time?

‘‘Takt time’’ refers to the rate at which customersare buying products from the production line; i.e.,the unit production rate that is needed to match

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236228

customer requirements. It is calculated by dividingthe total available time per day by the dailycustomer demand. The throughput required forthe annealed products is an average of 76,500 tonsper month. Assuming 30 days per month, theaverage daily requirement is thus 2550 tons perday. With an average coil weight of 20 tons, thistranslates into approximately 127 coils per day.ABS continuously runs three shifts per day, whichtranslates to 1440 working minutes per day, so thatthe takt time is thus approximately ð1440=127Þ ¼11:3 min per coil.

Question 2. Will production be directly to shippingor to a finished goods supermarket?

A ‘‘supermarket’’ is nothing more than a bufferor storage area located at the end of the productionprocess for products that are ready to be shipped(Rother and Shook, 1999). On the other hand,producing directly to shipping means that only theunits that are ready to be shipped are produced.Currently ABS produces all the annealed productsand sends them to a holding area where they arestored with other products waiting to be shipped.However, this is done based on a push system, andcoils of steel can wait a long time in this area beforebeing shipped. Our recommendation was that ABSproduce to a supermarket (warehouse) and movethe coils based on a kanban system. Whenever thesupermarket inventory is below a certain level thiswould trigger the TM (the last production stage) toschedule the annealed products to replenish thesupermarket according to the pitch, which isaddressed in more detail under Question 7.

Question 3. Where will ABS need to use pull systemsupermarkets inside the value stream?

The hot end at ABS is a continuous flow processby design, so that a supermarket at this end does notmake sense. The introduction of supermarkets isnecessary only at the finishing end where largeamounts of inventory exist between different work-stations. In addition to the shipping supermarketrecommended in Question 2, six additional super-markets are needed to create a continuous flow atthe finishing mill (cold end): one before the picklingline, one before the CR process, one before each ofthe three annealing processes (HBA, OCA, CA),and one before the TM (the reader is referred to theflow displayed in Fig. 1). Once a shipment of coils iswithdrawn from the shipping supermarket, thecorresponding kanban is sent to the TM where it

is placed in a load-leveling (heijunka) box. This inturn, triggers the production and movement ofmaterial from the earlier stages as described below.

The first supermarket recommended is ahead ofthe pickling area after the HSM. The latter currentlypushes coils to pickling, which causes inventory toaccumulate in two lines in front of the latter. Bothof these lines are shared resources (i.e., otherproducts can use them), and a kanban pull systemwas recommended to regulate the replenishment ofthis supermarket. A pull signal from the shippingarea is eventually transmitted to the HSM toreplenish the supermarket in front of the picklingarea, whenever the number of coils in the latterdrops to a trigger point.

The second supermarket is recommended tostabilize the CR process for the annealed products.The inventory after pickling and before CR is largeand both workstations are shared resources. Also,ABS runs its schedule in batches according to coilwidth, gauge, and product, so that it is necessary toset up a supermarket to accommodate schedulechanges. Again, a kanban pull system can be used toregulate the replenishment of this supermarket.Note that whenever this supermarket is full, thepickling process could run other products (non-annealed products) so that it is not idle. Also,pickling will no longer receive a schedule frombusiness planning for the annealed products.

The third, fourth, and fifth supermarkets arerecommended after CR and ahead of each of thethree annealing workstations. Thus the supermarketin front of the HBA process will be used for coils thatare ready to be placed in the HBA furnaces. The samething will apply for the supermarkets ahead of CAand OCA. Once again, the CR mill that suppliesannealing will no longer need to receive a schedule forthe annealed products from business planning and canrun other products types when those supermarkets areat their capacities. The last recommended supermarketis ahead of the TM. Since 96% of the products that goto the TM come from annealing, this supermarketarea can for all intents, be completely dedicated to theannealed products.

The kanbans at the supermarkets follow thestandard rules of a pull system. For example, thepickle line (supplier) is allowed to process the nextcoil in line as long as there is an empty coil spot inthe supermarket for the coil before CR (customer).By definition, if the supermarket is at its capacitythen this means that CR does not need another coil.In this case there are two things that can be done;

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 229

either the pickle line can slow its production rate tomatch that of CR or it should be halted. The firstoption is costly in a steel mill, while the second isalmost always infeasible. Therefore if the super-market at the pickle line is full our recommendationis that it be switched to satisfy other product typesuntil the time of the next order for the annealedproduct is reached. In doing so we preventproducing more than the capacity of the super-market and also satisfy requirements for otherproduct types, while avoiding shutting down thepickle line.

In the following questions we will address how aproduction order will be released and the timeincrement at which those orders will be released.

Question 4. Where can continuous flow be used?

Manufacturing assets in the steel industry aresuch that they cannot easily be moved into theclassical cellular arrangement, and batch sizes aretypically fixed. However, the steel industry doeshave a significant amount of continuous flowmanufacturing at the hot end. For example, startingfrom the blast furnace through the BOP, thedegasser/LMF, and finally the continuous caster,the flow is continuous since the liquid steel moves ina ladle in a batch size of one. At the finishing millhowever, the slab can move through one of manypossible routings using expensive general purposereduction equipment, which precludes cellular flow.Different CT and down times of the workstationsalso make it difficult to introduce continuous flow,and many of the workstations are restricted todifferent schedules depending on width, gauge andproduct type, so that it is unrealistic to join theseworkstations at the finishing mill to obtain acontinuous flow. Therefore, the focus at the coldend should be on developing a system to enable pullby the customer, rather than continuous flow. Inmost steel mills, the hot end (liquid steel) and thefinishing mill (solid steel) are located in the samearea; however, at ABS the two are actually aboutnine miles apart, purely based on historical circum-stances. This actually enables a natural decouplingof the two phases and allows a pull system to beincorporated.

The introduction of supermarkets that are con-trolled by a kanban system forces the entire cold endto pace every workstation to the speed of thebottleneck, which as the current state map indicatesis between the pickling line and the CR mill. Thusthe mill begins to take on the characteristics of an

assembly line where every product starts to flowrather than stop and start.

Question 5. What single point in the productionchain (the ‘‘pacemaker’’ process) should ABSschedule?

To stop overproduction at any workstation in thevalue stream, only one point in the supplier-to-customer value stream needs to be scheduled. Thispoint is called the pacemaker process, because thispoint sets the pace of production for all theupstream processes and ties the downstream andupstream processes together. Every workstationupstream produces by a pull signal from the nextdownstream process and flows downstream fromthe pacemaker must occur in a continuous manner.The pacemaker is typically the continuous flowprocess that is farthest downstream in the valuestream, so there should be no supermarket (otherthan finished goods) downstream of it (Rother andShook, 1999).

For ABS, since the hot end is located in adifferent facility than the finishing mill, the schedul-ing of a single process is unrealistic. For this reasonone schedule will be released to the continuouscaster to set the base for the hot end productionarea and the pacemaker process for the finishingmill is the TM; this is the final process and sets thebase for the entire production at the finishing mill.

Question 6. How should ABS level the productionat the pacemaker process?

The basis for addressing this question is todistribute the production of the three annealingprocesses uniformly over the production time at thepacemaker process. This means that several batchesof the same sequence must be scheduled. This willallow ABS to avoid long lead-times, large amountsof in-process and finished goods inventory, qualityproblems, and in general, help them avoid wastesrelated to overproduction. We consolidate thescheduling width and gauge for the coils so thatwe deal with only three different products; adjust-ment for width and gauge within a particular typecan be made as required. The key idea is for ABS tosend a schedule to the pacemaker process (TM) thatensures that each product is produced at a constantrate. We use a simple formula (Monden, 1998) todetermine the product sequence that levels the mixand has a constant rate for the three different

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236230

products:

dij ¼ ðj � 0:5Þ � ðT=DiÞ i ¼ 1; 2; . . . ; n and

j ¼ 1; 2; . . . ;Di,

where n is the number of different products to bemade, Di the number of units demanded per day forproduct i. T ¼ D1 þD2 þ � � � þDn is the totalnumber of units of all products to be made eachday, j the index for the job (unit) of product i, dij theideal position index for job (unit) j of product i inthe overall sequence.

For our case n ¼ 3, while the Di values are: 98, 14and 15 for HBA, OCA, and CA, respectively. ThusT is equal to 127. Ordering these jobs according todij sorted (shown in Table 1) one can see a patternstart to develop, yielding the following approximatesequence for smooth production (HBA-HBA-HBA-CA-HBA-OCA-HBA-HBA-HBA), (HBA-HBA-HBA-CA-HBA-OCA-HBA-HBA-HBA),yetc. We couldjust simplify the sequence to (HBA-HBA-HBA-HBA-HBA-HBA-HBA-CA-OCA).

Question 7. What increment of work (the ‘‘pitch’’)will be consistently released to the pacemakerprocess?

Depending on the sequence determined by thelast question, how often should we release andwithdraw (the ‘‘pitch’’) the increment of production

Table 1

Position index calculation for annealed products

Product ðiÞ Unit ðjÞ dij dij (sorted) Product-unit

HBA 1 0.648 0.648 HBA – 1

2 1.944 1.944 HBA – 2

3 3.240 3.240 HBA – 3

4 4.536 4.233 CA – 1

5 5.832 4.536 HBA – 4

6 7.128 4.536 OCA – 1

7 8.423 5.832 HBA – 5

8 9.719 7.128 HBA – 6

9 11.015 8.423 HBA – 7

10 12.311 9.719 HBA – 8

11 13.607 11.015 HBA – 9

12 14.903 12.311 HBA – 10

13 16.199 12.700 CA – 2

14 17.495 13.607 HBA – 11

13.607 OCA – 2

OCA 1 4.536 14.903 HBA – 12

2 13.607 16.199 HBA – 13

17.495 HBA – 14

CA 1 4.233

2 12.700

from the pacemaker process? The pitch is the basictime unit of the production schedule for a productfamily. In other words, it is the material transferinterval at the pacemaker process. The pitch iscalculated by multiplying the takt time by thefinished-goods transfer quantity at the pacemakerprocess. Since there is no container size involved inthe steel industry (we move one coil at a time), thenumber of kanbans will be the same as the currentdaily demand for OCA and CA. However onekanban will correspond to seven coils for HBA.Table 2 shows the number of kanbans required.Given a takt time of 11.3min, and considering thatthe transfer lot size is nine coils, the pitch isapproximately 1:40 h. Thus ABS will perform pacedrelease of work instructions and a paced withdrawalof finished goods at the TM according to this pitch.This means that the material handler will arrive atthe TM, remove the required kanbans from theheijunka (or load leveling) box of the TM corre-sponding to the next increment of work, and movethe coils just finished from the previous pitch to theshipping area supermarket. The number of pitchesrequired for every product is calculated as the dailyrequirement for every product divided by thetransfer quantity, while the time interval requiredfor every product to remove each kanban from theheijunka box is calculated by dividing the availabledaily time by the number of pitches for everyproduct (Table 3). The heijunka box is thus dividedinto 14 columns, each equivalent to about 1:40 hthat represent the frequency of introducing thekanban (work increment) to the TM. The columnfor each pitch interval will have three rows ofkanban slots—one for each of the annealedproducts.

Question 8. What process improvement will beneeded to achieve the future state design?

In order to accomplish the material and informa-tion flow envisioned by ABS, improvement andactions must take place to implement the future

Table 2

Number of kanbans required by product

Product Daily demand

(coils)

Transfer lot

size (coils)

Required number

of kanbans

HBA 98 7 14

OCA 14 1 14

CA 15 1 15

ARTICLE IN PRESS

Table 3

Number of pitches and material transfer times

Product Pitches per day Material transfer time

HBA 98=7 ¼ 14 1440=14 ¼ 102 min

OCA 14=1 ¼ 14 1440=14 ¼ 102 min

CA 15=1 ¼ 15 1440=15 ¼ 96 min

F.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 231

state. It is unrealistic to expect to obtain the benefitsof the supermarkets, kanban control, takt time, thepitch, production leveling, continuous improve-ment, and other changes discussed in the previousquestions without process improvement steps invol-ving specific lean tools; these are described in thenext section.

5. Tools for process improvement

Since our goal was to identify the potentialdynamic gains from implementing lean and todevelop a desirable future state map, we focusedon three lean manufacturing techniques that can bequantified and modeled objectively: a modified pull-type production system, setup reduction and totalproductive maintenance (TPM). To analyze andevaluate different scenarios for the future state map,a full factorial experimental design was planned forthe simulation, with the three factors being thetechniques just mentioned. Two levels were selectedfor each factor, thus resulting in 23 distinctcombinations for each replicate. These factors arenow discussed further.

5.1. Production system

A push system and a hybrid push–pull system arethe two levels for comparison that are used for theproduction system factor. The push system repre-sents the current situation at ABS where coils arepushed through the system. The hybrid systemhowever, is designed with the future state map inmind. In this system work will continue to bepushed through the hot end. However, the cold(finishing) end employs a pull system, starting withthe HSM at the beginning of this subsystem. Fromthe buffer area between the hot mill and the picklingline and all the way to the shipping area, the systemwill be based on a kanban pull system where theannealed products will be pulled from upstreamworkstations. The junction between the hot mill andthe pickling line thus forms the push–pull boundary.

In the simulation model, the portion of the systemup to the HSM is the same as the current one atABS. Starting with the pickling line the system wasmodeled as a pull system using kanbans to controlthe inventory between the workstations. This isdone by modeling each kanban between a pair ofworkstations as a resource. An arriving entity seizesone kanban and one workstation at the same time.As soon as the workstation finishes processingthe entity, the workstation is released; however,the kanban is retained. The entity then proceeds tothe next workstation. At this point the entity seizesthe workstation and a new kanban from the kanbanset for this latter workstation, while simultaneouslyreleasing the kanban from the previous workstation.Thus a kanban from one workstation is held untilthe entity receives a kanban from the subsequentworkstation. This ensures that the former does notbegin work until it gets a pull signal from the latter.In other words, the part retains the kanban from theformer workstation until it receives the next kanbanauthorization movement to the following work-station (Marek et al., 2001).

At the pull side of the hybrid system the totalWIP is limited to the sum of the number of kanbancards across each kanban set, where the latter isrepresented by a supermarket as defined in Section4. Since each coil in the supermarket has a kanbancard attached to it, the average system WIP levelmay be found by calculating the sum of the averageutilizations of the kanban resources in the simula-tion. The comparison of WIP inventory for the pushand the hybrid system is based only on theinventory ahead of the pickle line and downstreamto the TM. The reason for this is that the differencebetween the two systems in terms of WIP inventorywill be after the push–pull boundary point; allearlier inventory levels are identical since thesystems being compared are identical up to thispoint.

5.2. Total productive maintenance

The two levels for the TPM factor are labeled‘‘without’’ and ‘‘with.’’ The former identifies currentmaintenance procedures followed by ABS, while thelatter models a proposed TPM procedure that splitsthe same scheduled maintenance time into smallerincrements, i.e., it separates the maintenance pro-cess into smaller portions that are performed morefrequently. Maintenance is planned in such a waythat it cascades through the process so that the

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236232

inventory shortages created by work stoppage formaintenance result in minimal disruption of flow.As an example, when the pickle line is maintainedthen the kanbans in its supermarket (the area aheadof cold rolling) would empty; therefore, the nextmaintenance operation is scheduled on the cold mill.This permits the pickle line to restock its super-market, and so on.

TPM can significantly reduce random machinebreakdowns and in turn, inventory and lead-time. Itis usually defined in terms of an increase in overallequipment effectiveness (OEE), which in turn is afunction of down time and other production losses(Nakajima, 1989). Suehiro (1992) states that ma-chine breakdowns and minor stoppages account for20–30% of loss in OEE; Ljungberg (1998) alsoreports a 20% figure for the same. Volvo Gentreports that the OEE in the company increased from66% to 69% before implementing TPM to 90%after TPM where most of the increase is a result ofthe elimination of machine breakdowns and minorstoppages (Ljungberg, 1998). Similarly, Avon Cos-metics report significant increase of OEE after TPMwas implemented at its pump spray line (Ljungberg,1998). Based on this experience we assume a 20%increase in OEE at ABS.

Table 4 shows proposed TPM times at thefinishing mill. The maintenance times were chosenbased on optimistic but reasonable estimates afterconversations with floor personnel. One issue thathad to be taken into consideration with theproposed TPM program is that the time for eachof the different maintenance tasks for a givenprocess should not exceed the total proposed(reduced) maintenance downtime. This issue wasdiscussed with ABS and it was confirmed that theproposed downtime should be feasible.

Table 4

Proposed TPM times at finishing mill

Process Maintenance Day

Uptime (days) Downtime (min)

HSM 7 240 Monday

8400 Pickle 7 240 Tuesday

6400 Pickle 7 240 Wednesday

CRM 7 240 Thursday

TM 7 240 Friday

5.3. Setup time reduction

The two levels for the setup reduction factor arealso labeled ‘‘without’’ and ‘‘with.’’ The ‘‘without’’level models the current situation at ABS with setuptimes the same as they are now, while ‘‘with’’denotes reduced setup times. Again, the changeoverreduction times were selected based on optimisticbut reasonable estimates, with values that arerealistic for ABS to drive their CO time down.These were based on extensive discussions with floorpersonnel including operators and engineers and aresummarized in Table 5. For example, we analyze asetup time reduction for the HSM from 35 to 10minfor the backup rolls and 120–20min for the workrolls. For CR the reductions analyzed were from 15to 5min for the backup rolls and 120–20min for thework rolls. Other setup times reductions are asfound in Table 5.

6. The simulation model

To evaluate potential gains based on the im-plementation of the tools described in Section 5 andbased on the questions analyzed in Section 4, adetailed simulation model was developed usingSystem Modeling Corporation’s Arena 5 software.We began with a model for the current system,which was later modified to model the proposedfuture state. Before evaluating the future stateconsiderable effort was expended to verify andvalidate the model for the current system. Verifica-tion is the process that ensures that the simulationmodel mimics the real system (Law and Kelton,1991). Since this model is large with many types ofentities (grades and products) in the system,verification required that every kind of product betraced and checked to ensure that it follows itsrequired sequence. In order to see if the model

Table 5

Proposed setup reduction times at ABS

Process Setup Times (min)

Work Rolls Backup Rolls

Current Proposed Current Proposed

Hot strip mill 120 20 35 10

Pickling 15 5 – –

Cold reduction 120 20 15 5

Temper mill 90 20 7 5

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 233

represents the real system adequately, the first stepwas to check the code and verify the model logicand the experimental conditions; followed by acareful trace study where various entities weretraced from the point of creation until the point ofdisposal from the system. Finally, a detailedanimation was used to further verify that the modelsufficiently replicated the real system.

Validation of the model calls for comparingoutputs of the simulation to those from the actualsystem. Measures that we included were inventoryat the finishing mill and the total time in the system,for both of which actual data was available. Thesimulation model was run for a one-year period,which is equivalent to an expected 11,520 heats(furnace batches) out of the BOP, so that the modelcan be validated when it is in steady state. Table 6shows the actual values and the simulation resultsthat were obtained by running the model. It shouldbe noted that the figures represent average values.From the table it is clear that the numerical outputsfrom the simulation are all within the range of theactual data.

Our simulation is non-terminating (Law andKelton, 1991) but initial conditions do influencethe initial dynamics of the system. Starting with anempty system at time zero, a transient (warm up)period was used for the system to load itself withentities and subsequently reach steady state. Thewarm up period for our simulation model wasestablished by carrying out five replications witheach having a run length of 1 year (11,520 heatsfrom the BOP). The five replications examinedsuccessive observations of various performancemeasures. For example, Fig. 2 shows the plot forthe total work in process inventory in the system asa function of time. Based on this type of analysis, awarm up period of 60,000 minutes (42 days) wasadopted.

Table 6

Performance measures: actual vs. simulation

Performance measure Actual range Simulation

Entity lead-time [30–49 days] 34 days

Hot strip mill inventory [1000–5000] 3703 slabs

Cold mill inventory [250–2000] 1755 coils

HBA inventory [250–1750] 620 coils

CA inventory [100–750] 121 coils

OCA inventory [100–750] 636 coils

Temper mill inventory [150–750] 653 coils

Number of coils per month [9000–9800] 9466 coils

7. Simulation results and assessment

Once the simulation model for the current systemwas verified and validated it was used to evaluatethe future state map and assess the relative impactof adopting the lean approach detailed in theprevious two sections. It is worth mentioning thatthere are other lean techniques like 5S and visualsystems, the benefits from which are not directlyquantifiable and cannot be modeled as part of asimulation model. These have therefore not beenincluded in our analysis, but these techniques couldeasily be applied at numerous places within theproduction system at ABS, and can be expected tofurther increase the potential gains from theadoption of lean.

Based upon our initial observations from thecurrent state map and discussions with managers atABS it was decided that two primary performancemeasures would be examined: production lead-timeand work-in-process inventory. The latter is eval-uated as the sum of the WIP starting at the picklingline and ending at the TM; only this portion of theWIP is considered because the systems are identicalup to the push-pull boundary point at the picklingline. As mentioned earlier for the hybrid productionsystem the WIP inventory is just the sum of theaverage utilizations of the kanban resources.

Two sets of factorial design experiments were ranto study the effect of the three factors on theproduction lead-time and WIP inventory, respec-tively. Each factor had two levels, and for each level-factor combination the experiment is replicated fivetimes using the simulation model and is completelyrandomized. Thus, eight simulation runs were carriedout, each with five different replications. Analysis ofvariance (ANOVA) was used to formally study theresults and determine the significance and magnitudeof all effects and interactions. The statistical analysiswas done using Minitab with outputs displayed inTable 7. The p-values indicate that for lead-time theproduction system and TPM are significant, whilesetup reduction, and all two- and three-way interac-tions are not.

For the WIP inventory, once again the maineffects of the production system and TPM aresignificant. Interestingly, the table shows that thetwo-way interaction between the production systemand TPM is also significant here. To betterunderstand this interaction, Fig. 3 presents a plotof the production system-TPM interaction, thesignificance of which is indicated by the lack of

ARTICLE IN PRESS

Table 7

p-Values for effects

Term p-value

Lead-Time Inventory

Constant 0.000 0.000

Prod Sys 0.000 0.000

TPM 0.000 0.000

Setup Red 0.815 0.815

Prod Sys � TPM 0.000 0.000

Prod Sys � Setup Red 0.632 0.632

TPM � Setup Red 0.783 0.783

Prod Sys � TPM � Setup 0.815 0.815

Hybrid

Push

90

80

70

60

50

40

30

20

10

WithTPM

Prod System

Mea

n

Interaction Plot (data means) for Inventory

WithoutTPM

Fig. 3. Main effect and interaction plot for inventory.

10

10

9

9

8

8

7

7

6

6

5

5

4

4

3

3

2

2

1

1

0

Avg

WIP

(X

103 )

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

Simulation Time (X103)

Transient Period

Fig. 2. Transient period analysis for the average WIP inventory for five replications.

F.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236234

parallelism of the lines. In the figure, the lower solidline represents the hybrid production system and theupper dashed line represents the push system. Theinterpretation of the interaction graph is that goingfrom no TPM to TPM when the production systemis a hybrid will not change the level of WIP

inventory, whereas going from no TPM to TPMwhen the production system is a push system willdecrease the level of WIP inventory. An intuitiveexplanation for this is that the WIP inventory in thepull system is dependent on the number of kanbancards that is predetermined before the run. This

ARTICLE IN PRESSF.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236 235

makes the change in WIP inventory for the hybridproduction system (WIP inventory is the sum of theaverage utilization of the kanbans, which aremodeled as resources) insignificant when usingTPM. Even though TPM was found to besignificant, the kanban pull system is so instru-mental in reducing the WIP inventory that the effectof TPM is relatively small.

Based upon our the simulation experiment theactual magnitudes of the improvements in the twoselected performance measures were significant. Theresults indicate that using a hybrid productionsystem and TPM could potentially reduce the totalproduction lead-time from its current value of 48days to under 15 days, a reduction of almost 70%.Setup reductions in the amounts listed in Table 5does not seem to have any significant additionaleffect on lead-time. In terms of the WIP inventory,the experiment revealed that the new system couldpotentially drive down the current average inven-tory level across all stations between the pickle lineand the TM from the current value of about 96 coilsto around 10 coils; a reduction of almost 90%.

In summary, the results indicate that a hybridproduction system with TPM can have enormous

Fig. 4. Future s

effects on reducing both lead-times and WIPinventory; for this particular instance setup reduc-tion did not have a similar effect. However, thisdoes not necessarily mean that it is not a valuablelean tool for ABS. Rather, the effect of the hybridsystem and TPM outweigh the advantages of setupreduction in this particular case.

8. The future state map revisited

The future state map for the annealed product forABS is shown in Fig. 4. The results of the foregoinganalysis are documented on the future state map,and the proposed lean tools are shown as kaizenbursts to highlight the improvement areas. Alsoshown are the supermarkets between each processafter the HSM. As we can see in the map, ABSreceives two schedules only; one at the continuouscaster for the push system at the hot end and theother one at the TM for the pull system at thefinishing end. With the new improvements at ABSthe value added time (5 days) is up from approxi-mately one-tenth of the production lead-time in theold system, to approximately one-third of the totalproduction lead-time of slightly under 15 days

tate map.

ARTICLE IN PRESS

Table 8

Assessment of lean tools in the steel industry

Lean tool Applicability

Cellular manufacturing Probably inapplicable

Setup reduction Partially applicable

5S Universally applicable

Value stream mapping Universally applicable

Just-in-time Partially applicable

Production leveling Partially applicable

Total productive maintenance Partially applicable

Visual systems Universally applicable

F.A. Abdulmalek, J. Rajgopal / Int. J. Production Economics 107 (2007) 223–236236

(12.84 in waiting plus about 2 in processing). Putanother way, non-value added time is about 8.6times the value-added time according to the currentstate map, but with the future state map the value ofthis multiplier drops to about 2.

9. Summary

Applications of lean manufacturing have beenless common in the process sector, in part becauseof a perception that this sector is less amenable tomany lean techniques, and in part because of thelack of documented applications; this has causedmanagers to be reluctant to commit to theimprovement program. This paper takes a case-based approach to address both issues. Manyindustries in the process sector actually have acombination of continuous and discrete elements,and it is in fact quite feasible to judiciously adaptlean techniques. We demonstrate this with the steelindustry where several lean techniques can besuitably adapted. Table 8 summarizes this conten-tion. Furthermore, for managers who might beconsidering implementing lean manufacturing butare uncertain about the potential outcomes, wedemonstrate that a detailed simulation model can beused to evaluate basic performance measures andanalyze system configurations. The availability ofthe information provided by the simulation canfacilitate and validate the decision to implementlean manufacturing and can also motivate theorganization during the actual implementation inorder to obtain the desired results.

References

Abdelmalek, F., Rajgopal, J., Needy, K.L., 2006. A classification

model for the process industry to guide the implementation of

lean. Engineering Management Journal 18 (1), 15–25.

Abdullah, F., Rajgopal, J., 2003. Lean manufacturing in the

process industry. Proceedings of the IIE Research Confer-

ence, CD-ROM, Portland, OR, IIE, Norcross, GA.

Abdullah, F., Rajgopal, J., Needy, K.L., 2002. A taxonomy of

the process industry with a view to lean manufacturing.

Proceedings of the American Society for Engineering

Management, Tampa, FL, pp. 314–321.

Ahmad, M., Dhafr, N., Benson, R., Burgess, B., 2005. Model for

establishing theoretical targets at the shop floor level in

specialty chemicals manufacturing organizations. Robotics

and Computer-Integrated Manufacturing 21 (4–5), 291–400.

Billesbach, J.T., 1994. Applying lean production principles to a

process facility. Production and Inventory Management

Journal, Third Quarter 40–44.

Cook, R.C., Rogowski, R.A., 1996. ‘‘Applying JIT principles

to continuous process manufacturing supply chains. Produc-

tion and Inventory Management Journal, First Quarter

12–16.

Detty, R.B., Yingling, J.C., 2000. Quantifying benefits of

conversion to lean manufacturing with discrete event simula-

tion: a case study. International Journal of Production

Research 38 (2), 429–445.

Feld, W.M., 2000. Lean Manufacturing: Tools, Techniques, and

How To Use Them. The St. Lucie Press, London.

Law, A.M., Kelton, W.D., 1991. Simulation Modeling and

Analysis, second ed. McGraw-Hill, New York.

Ljungberg, O., 1998. Measurement of overall equipment

effectiveness as a basis for TPM activities. International

Journal of Operation and Production Management 18 (5),

495–507.

Marek, R., Elkins, D.A., Smith, D.R., 2001. Understanding the

fundamentals of Kanban and CONWIP pull systems using

simulation. Proceedings of the 2001 Winter Simulation

Conference, pp. 921–929.

McDonald, T., Van Aken, E.M., Rentes, A.F., 2002. Utilizing

simulation to enhance value stream mapping: a manufactur-

ing case application. International Journal of Logistics:

Research and Applications 5 (2), 213–232.

Melton, T., 2005. The benefits of lean manufacturing: what lean

thinking has to offer the process industries. Chemical

Engineering Research and Design 83 (A6), 662–673.

Monden, Y., 1998. Toyota Production System—An Integrated

Approach to Just-In-Time, third ed. Engineering & Manage-

ment Press, Norcross, Georgia.

Nahmias, S., 2001. Production and Operations Analysis, fourth

ed. McGraw Hill, New York.

Nakajima, S., 1989. TPM Development Program: Implementing

Total Productive Maintenance. Productivity Press, Cam-

bridge, MA.

Radnor, Z., 2000. Changing to a lean organisation: the case of a

chemicals company. International Journal of Manufacturing

Technology and Management 1 (4/5), 444–454.

Rother, M., Shook, J., 1999. Learning to See: Value Stream

Mapping to Add Value and Eliminate Muda. The Lean

Enterprise Institute, Inc., Brookline, MA.

Suehiro, K., 1992. Eliminating Minor Stoppages on Automated

Lines. Productivity Press, Cambridge, MA.

Womack, J.P., Jones, D.T., Ross, D., 1990, The Machine that

Changed the World. Macmillan Publishing Company,

Canada.