Management of WIP Inventory in JIT Environment Under...
Transcript of Management of WIP Inventory in JIT Environment Under...
Journal of Scientific & Industrial Research Vol. 60, March 200 I, pp 256-263
Management of WIP Inventory in JIT Environment Under Cellular Layout - A Case Study
S Venkataramanaiah·, K Krishnaiah and R Pichandi t
Industrial Engine ring Division, Department of Management Studies, College of Engineering, Guindy,
Anna University, Chennai - 600 025, India, E-mail: [email protected]
In order to enhance the productivity and quality of products and services under the stiff competition in the global market, organ isations are compelled to change their manufacturing technologies and strategies. One such strategy is to change the traditional layout into a cellular one to retain tlexibility of jobs hop and to bring in more related benefits. This paper discusses the effect of work- in-process (WIP) inventory on throughput rate in a cycle manufacturing company which switched over to a cellular manufacturing recently. A GPSS simulation model was developed to study the performance of the production system. For various models of the cycles, past data on demand, demand arrival time and process times were gathered from each cell. The distributions of these data were used and simulation was carried out for various combinations of WIP inventory with throughput rate as the performance measure. Finally WIP inventory levels for each cell were suggested for future production. Suggestions were also given for effective implementation of JIT strategies under cellu lar layout to improve the
overall productivity. Indicators of a manufacturing system, where the technology was stable and the market structure was static and pre-defined, could be constructed upon the degree of production economy achieved . Normally, these are achieved through incremental innovations. Throughput, defined as the ratio of total sati sfi ed demand to the total generated demand, and expressed as percent customer orders fulfilled, could be considered as an indicator of a stabilised manufacturing system. A related indicator that is somewhat dependent on the technology in operation was the ratio of WIP inventory levels at various cells/workstations to the throughput rate of the system, under thmughput maximisation. The distribution of WIP along the ce ll s and not the total WIP, was the most important process indicator under throughput maximisation.
Introduction The core of Japanese productivity improvement in
repetitive manufacturing is Just in time (111') philosophy. 111' concept proved effective in a pull production system that pulls material from a prior process in support of the final assembly. In conventional/pure push (Non 111') production system facilities are arranged according to their generic process characteri stics. Activities/jobs that require various resources wi ll visit according to sequence of operations and leave the system. In convent ional systems, planning takes place around some combination of forecasted demand and current actual customer orders. A production plan is developed that pushes through the manufacturing cycle and is genera lly less productive ' . In Cellular manufacturing, often referred to as the natural candidate for implementation of 111' and TQM, resources that are required to produce a product or set of s imil ar products are arranged in a cell according to manufacturing requirements of parts . Cellular layout helps in achiev-
* Addrcss for correspondance tTata Consultancy Services Ltd , Bangalore
ing the goals of 111' and TQM2. In 11T production system kanban cards are used to withdraw and produce the required parts/subassemblies' . The objective of this paper is to design suitable work-in-process (WIP) inventory levels to maximise throughput rate/customer service in a cycle manufacturing company where 11T concept has been followed under cellular configuration.
Literature Review Kanban based operational planning and control is
sues have been tackled in a number of stud iesl.H> built upon analytical and/or simulation modelling. Schroer et ai. 5, presented a microcomputer simulation stud y of 11T with Kanban based manufacturing system. It developed a simu lation model using 'SIMAN ' of a simplified 11T manufacturing system using one and two card kanban to vary the level of in process inventory in comparison with production rates. The work concerned with minimisation ofWIP inventory at variou s
1
VENKATARAMANAIAH el at.: lIT ENVIRONMENT UNDER CELLULAR LAYOUT 257
stock points in the system and to balance the system in terms of resource utili sation and queues. Mejabi\ described application of simulation constructs for JIT
modelling and proposed a new language construct based on kanban sati sfaction paradi gm to provide required features for JIT simulation mode l for integrat
ing new constructs into SIMAN language. Berkley" discussed issues regarding minimum performance leve ls for Kanban controlled I ines and methods to find minimum performance levels necessary to guarantee production rates independently by using average station process ing time. Yavuz and Satirfi , presented a Kanban based simulation study of a mixed model JIT manufacturing line and features of a hypothet ica l manufacturing line in terms of general structure, major components and operational characteri stics. Researchers used simulation languages such as SIMAN, and SLAM to model the system and help in dec ision making. Since the organisation where the study was carried out was in the early stages of implementati on of JIT strategies, a simple GPSS simulation mode l has been developed. WIP inventory levels at various stages in the production system have been des igned and studied its effect on system performance, where JIT philosophy is followed under cellular confi gurati on.
Description of the production system The flo w of materi al in the production system is
shown in Figure I . The system consisted ten manufacturing ce ll s arranged in a line type of layout where JIT concept was followed. There were three production lines name ly Frame, Fork and Mudguard , processed in paralle l. Parts from these three production lines were g rouped into a set of defined quantities at Kitting cell (# I 0). There were three major stages in the system, viz, kitting, finishing and phosphating. The object ive was to meet the market demand with minimum WIP and finished goods inventory (FGI) at various stages in the syste m apart from the in ventory at the raw materi al stage. As kitting cell was the las t ce ll in production syste m, it acts as schedule dri ving cell for other cells. Bins and racks were used to transfer material from one cell to another. Whenever there was a demand, equi va lent production was tri ggered. The kitting cell checked for required quantity and means to meet the generated demand. In case of shortage of required quantity at kitting ce ll , it gave signa l to the previous cell and pulled required quantity if enough WIP inventory was ava ilable otherwi se, that centre triggered its production . Other ce ll s in the system, fo llowed the same procedure .
Problem Definition JIT production system helps increase inventory turn
over and reduce equipment breakdowns, tool failures, problems in production planning and control. Excess inventory is an evil in JIT philosophy. Optimal or near optimal results can be effectively achieved by
designing proper WIP inventory levels at appropriate stages in a system . A max imum total WIP in the sys
tem was no longer the goal. It was therefore necessary to evaluate the system by considering WIP invento ry levels at all cells instead of treating the entire system as a single entity. The key point in JIT production system was that the system had to produce when there was a demand ie, the generated demand should be satisfi ed by the system. Throughput, ie, percent customer orders fulfilled (-Ratio of total sati sfi ed de mand to
the total generated demand) was considered a su itable measure of performance to evaluate the JIT producti on system for various combinations of WIP in ventory levels. Hence, the objective of the study was to set up suitable WIP inventory levels at variou s cell s/ workstations such that throughput rate of the syste m was maximized .
Inter alia, indicators of a manufacturing system, with the stable technology, static and pre-defined, market
structure could be constructed upon the degree of production economy achieved . These were norm a ll y achieved through incremental innovati ons. Implementation of a JIT, with the objective of a minimi sed WIP inventory in a cellular production environment, came under the incre mental changes. The changes through incremental innovation that could be achieved percentage included cost reduction, closer and deeper market-orientation of the manufacturing, amenability of the production to the demand fluctuation and fin all y, competition orientation. Throughput, was cons idered
as a good indicator of a stabi li sed manufacturing system. A related indicator, somewhat dependent on the tec hnology in operation , and hence whose use should remain limited to the benc hmarking, was the rati o of WIP inventory levels at various ce ll s/workstati ons to the throughput rate of the system, with a v iew of throughput maximisation . As claimed in this paper, it was the distribution of WIP along the ce ll s and not the total WIP, that was the most important process indicator under the aim of throughput max imisati on. These two indicators, could benchmark if limited to stable technology and competing manufacturing firm s. Most important perhaps is the fact that these are process indicators . Most other indicators were e ithe r input or the output types. The suggested indicators referred to
258 J SCIIND RES VOL 60 MARCH 2001
CalCJJE t.c ~o. of lots [(} rm..xl~Il;~{l\' )
r"\' t I -~----'I
J'un !'n;ori! {I!J: '(:- :t ;m,j I I
fllr .... ,ud Xl ,;"~ (:!,,,,: IIII l!. r~11I !!l' ~ml updz.:e \'(1 P
)J(l. klI1i l]C l!f'!!fZll:..:o t----~-~"""'I
I I
t'lgurc I - IVlalenal [lOW magram - proauc[Jon SYSlem
the process internal systems and, therefore, were more suggestive of the dynamics of process parameters .
Components of the system modeled
Following are the some of the important components of the system modeled.
Products - The system was capable of producing six (n=6) types of products (cycle), each type of product
was handled in the system with Kanban (container) size of 200 units. Work Cells - The production line was composed of ten (k= I 0) work cells. The line was capable of producing six (n=6) type of products. Each work cell had a processing capacity of one batch (size 200) at a time. Each cell in the system was designed by considering the processing requirements of various models that were produced . There were three main lines for frame, fork and mudguard , as shown in Figure I .
YENKATARAMANA IAH 1'1 ((/.: JIT E YIRONMENT UNDER CELLULAR LAYOUT 2'i<)
Work in p rocess (WI? ) In venlory - Any job in a queue, moving from one operati on to the nex t, being delayed for some reason, being processed, or residing in the system or sub assembly inventori es was cons idered as work-i n-process (WIP) in ventory, a special case of transit inventory. Each ce ll was equipped wi th both inbound and outbound stock ing po in ts . Storage space, and inventory carry in g cost were considered as constraints on WIP inventory.
Fin ished Coods In ventorv (FCI) - Each cycle had its own FGl. The final cell began it s production to rep leni sh the consumed containers of fi ni shed goods. The removaf' of a container from its WI P in ve nt ory tri ggered the manufacturing of that product. Hence, unless there was a demand, no prod ucti on was init iated . External demand was satis fied frol11 the in ventories. If there was a dearth of sufficient number of items in FG f, the customer waited until the required amoun t was accumu lated and it was measured as customer service which was strongly rel ated with WIP and FG in ventory in the system.
Operational Characteristics of the Model I mportant operational characteristics of the system
are given as follows:
• The system could be viewed either as a component of a compos ite man ufacturing system, where producti on was carri ed out to satisfy the internal de mand of the succeed ing manu fac turin g cell , or a complete configuration in itself wh ich prod uced to satisfy ex ternal demand.
• Time between demand arri va l and demand sat isfac ti on were the random vari abl es which were generat ed from the appropri ate distribution , based on the past data for each product/model.
• Each model/product fo llowed the same processing route along the line.
• Prod uction was tri ggered only upon the generation of a demand for a given prod uct/model with no lost sales.
• For a given prod uct, the producti on lot size was 200 units at each ce ll and hence prod ucti on was carri ed out in batches. However. withdrawa l uf units from final assembl y was in si ngle un its.
• Set up was required at a wash cell when different types of products were processed sequentiall y, set up times are constant at each work cell.
• A ce ll may suffe r from breakdown. Time between breakdown and repa ir times were included in the process ing time.
• Processing at a station was carried out with no
defecti ves. Hence, a perfec t quality confo rmation was assumed along the line.
• Conti nuous supply of raw material s was assumed. Therefore the first ce ll in the line never stan·ed. Si mil arly continuous bought out components were assumed , therefore kitting cell never starved for bought out items.
Data Collection and Analysis Data, such as demand and demand arrival time wt:re
collected from the past records for vari ous model s (cycles) incl uding process ing time at eac h cell . Inter-arrival ti me of demand, frequency of each day, illl crarr ival frequency, and mean arri val time were calculated from the data. Mean demand, demand frequency/ dis tribution was calcul ated using the follow ing relation .
Let N be the sa mple size (No. of data collected), K he the nu mber of classes, ca Iculated usi ng Struge' s rule K = I + 3.3 loge N.
Le ngth of each class was fo und usin g lower and upper va lue (LV and UV)
Difference (D) = uv -LV Range R = D/K
Class interval is calcul ated based on the range. Frequency (F) and mi d value (X) of range i was calculated for the corresponding class in terval and thereby probability di stributi on of each class interval wa~ arri ved at. Mean demand was computed usin g the relation,
Mean demand = I F X / IF I I I
Simulation Model Development Simulati on model sought to duplicate behaviour of
the system under inves ti ga tion by study ing the int eraction among its components. The output of silllulaii on model was presented in terms of selected measures (WIP, throughput/service) that rdlected the perfo rmance of the system .
In the present study a GPSS model was builtlo design WIP in ventory levels and to its effect on the systemlhroughput was studi ed . GPSS maintained a simulati on c lock , schedul es events to occur in future simulated time, caused the events to occur in a proper timcordered sequence and assi gned relative priorities lI sed in resolving time ties. Production cont rol system fl ow chart is shown in Figure 2.
260 J SCIIND RES VOL 60 MARCH 2001
The motivation in using GPSS model was its ability to understand the best known blocks and standard output, etc . The model was run at different simulation run periods and appropri ate run length was selected based on simulated mean and actual mean for the collected data . Simulation run length used in the study was set as 30000 minutes of simulati on clock time.
Model Frame Description The following sections are used in the system for
controlling purpose: Inter arrival time section (lAT) , Demand generation section, Pull initiating section , and Satisfaction section .
fnter Arrival Tim e Seclion ( fAT) - The inter arri val time followed an ex ponential di stribution . fAT was calculated by the following relation .
Where IAT""11PIC stands for the sample inter arrival time, RN
j uniform random number generator, and loge
represents natural logarithm operator.
The log function was approx imated with continuous GPSS function called EXAD, obtained by multiplying FN$EXAD with fA T,IV ~ Thus the inter arrival time was generated using GENERATE Block.
Demand Generation Section - Based on the cuntulati ve probabi I ity of demand, an appropriate demand function EXDD was defined . Using function EXDD demand was generated.
Pull Initiating Section - Whenever demand was generated, it was converted into number of lots. Split block created transactions, eq uivalent to the number of lots required. Thus pul l was being initiated.
Demand Sati.llaction section - Demand was satisfi ed by checking the in ventory ava il ab le at kitting cell (# 10) or by drawing parts from preced ing cell s. Similarly other cells drew products from its precedin g cell. Thus, demand was sati sfied/customer order was fill ed.
GPSS Model Description (Working Mechanism)
The GPSS blocks used in the mode l were GENERATE, ADVANCE, SAVE, SPLIT, ASSIGN , TEST, TERMINATE, TABULATE, ENTER and LEAV E. On generation of a demand , ava il ability at ce ll # 10 will be checked if enough number of units were ava ilable and demand would be met, otherwise the requ ired quantity would be pulled from its preceding stage (fin-
ishing stage). Similar procedure would be followed at the other cells in the system. Once after di spatch of units, WIP at various cells would be updated by producing the quantity equivalent to the dispatched quantity (lots) . A GPSS flow chart for the system under consideration was drawn with kitting cell and fini shing stage along with frame line. Similar procedure was followed for mudguard and fork line.The step-wise procedure of GPSS model is given below:
GENERATE Block generated a transaction based on the average inter arrival time (IAT,,,,) and the GPSS inter arrival time function (FN$EXAD).
V$DMND generated the demand and saved In SAVEVALUE block.
TABULATE block tabulated the demand .
SAVEVALUE Block calculated the total demand (x l) by adding the demand generated in the previous Block.
At TEST block generated demand would be compared with the WIP at cell 10 and met if enough WIP was available. If WIP at cell was less, it would pull from its preceding cell (s). SAVEVALUE block saved the unfilled demand in a variable (UFD),
Variable V$DNK converted the demand into number of lots and stored remainder quantity in the remainder quantity variable, V$REMAIN .
SPLIT block propagated transact ions equivalent to the number of lots.
The ENTER Block acted as BIN for the WIP at the kitting cell. Whenever the Block was free, it allowed a transaction to enter in it ; otherwi se th e transacti on stayed in the previous block itself.
Once the transaction entered in the ENTER block, TEST block checked the WIP in all immed iate preceding cells . If enough WIP was available, kilting cell pulled the WIPs, completed the operati on and released the BIN. If the immediate previous ce ll WIP was not sufficient to satisfy the requirement of the kitting cell , previous cells pulled the WIPs from its immed iate previous cell s.
GENERATE block generated a transaction at a spec ified time to complete the simulation.
Results, Conclusions and Future Scope of Work Results - Simulation ex periments were conducted fo r different simulation run periods. Ba'se simulation run period was selected fro m the analys is by comparing simulated mean and actual mean. The model was simulated for the base simulation run period by keep ing zero WIP in ventory at all cell s. Similarly the WIP at
~"
VENKATARAMANAIAH el al.: JIT ENVIRONMENT UNDER CELLULAR LAYOUT 261
Stage-3 Stage-2 Stage-l Stage-O
RawMa terial iF
- I Phosphating & Frame-
~ Cell I -'" painting Finishing
Cell 2 Cell 3 ine Frame I
60 min/200 units 47 min 45 min
45 IT in
terial RawMa (Cycle s)
Mudguard Phosphating Mudguard- ~
Cell-4 & painting Finishing Kitting .. Cell 5 Cell-6 Cell-I
Mudguard Finished goods
line 40 min/200 units 28 min 45 min t
BI ut
ite s
Raw Material Fork Phosphating & Fork-
Cell 7 painting Finishing Fork line Cell 8 Cell 9
60 min/200 units 7 min 45 min
Figu re 2 - Production control system - flow chart
a ll ce lls were increased by 200 units, and repeated trials were conducted by increasing the WIP by 200 units equally at all ce lls . Throughput rate (per cent service) was calcul ated for these tria ls. From these trials, limit on WIP inventory was fo und to be 0 - 600 in steps of 200 and results of the tri als are given in
Table I .
Taking minimum WIP as 0 and max imum WIP as 600 units in steps of 200 units, various combinati ons of WIP were obtained using the model. S imulati on was run fo r 500 combinatio ns of W IP in ventory levels. Sample results a re g iven in Table 2 with assoc iated WIP in ventory levels. Three combinati ons gave above 99 per cent service leve l. WIP in ventory kept at diffe rent cells for combinati on A was 1800 units, fo r B 3000 units and for C 2800 units. Combinati on A gave the least in ventory to be kept and it was chosen as the better combination among A, B and C. Combination E resulted in the lowest throughput wi th hi gh WIP than combinations A, D and F. From the results ob ta ined , it is apparen t that the overall syste m performance was affected by the distributi on of W fP inventory at various stages in the system. Combination A gave the low-
est WIP w ith 99.78 per cent th roughput whereas combination F resulted in the lowest throughput of 57.62 per cent for the same WIP leve l. Combination B and C showed the same throughput of 99.90 per cent at d ifferent WIP levels. Alte rnati ve D and F showed the same throughput of 57 .62 per cent with WIP at 2000 and 1800, respecti vely.
Even though F and A had the same tota l W IP inventory ( 1800), combinati on A gave a bette r throug hput rate w ith service rate 99.78 per cent compared to the combinati on F with service rale 57.62 per cent. It showed that the distribution of WI P at various stages in the system was more important unlike the total WIP. Hence combination A is recommended as suitable WIP inventory at di fferent ce ll s in the system based on the performance measure used .
Conclusions
As the company changed its layout fro m trad itiona l (process) to cellul ar layout with lIT concept, the future production could be pl anned based on these g uide lines . It was believed that thi s ki nd of simple management tool would help the manu fac turing concerns to
262 J SCI I 0 RES VOL 60 MARCH 2001
Tab le I - Simulati on results for equal WIP level
T rial II W IP - in ventory at cc ll Throughp Total WIP
(x 100 Units) 1Il
(x 100
2 3 -+ 5 6 7 8 9 10 (per ce nt ) Unit s)
0 0 0 0 0 0 0 () 0 () 60.93 0
2 2 2 2 '2 2 2 2 2 2 2 9R.80 20
3 4 -+ 4 -+ 4 4 4 4 4 4 75.53 40
4 6 6 6 6 6 6 6 6 6 6 82.5 1 60
T able 2 - Saillp ic result s- simul .lli on stud y
Combination W IP - invcntory at Cell Throughput Total W IP
(x 100 Un its)
2 3 -+ 5 6
A 4 () 4 4 2 ()
B -+ () 6 '2 '2 6
C 0 -+ '2 '2 '2 6
0 2 -+ 4 0 () ()
E 2 0 -+ 0 () -+
F 4 2 2 () () 0
improve their overa ll productivity, significantly but incrementally.
The inJ icators, of a man ufac turing system where the tech nol ogy i ~ stable and the market structure is ~ta ti c and pre-defined , can be constructed upon the degree of product ion economy achi eved. Ordinarily, thest: are ach ieved through incremental innovations. II 1piementation of a JIT, with the objecti ve of a minimi.-ed WIP inventory in a cellular prod uct ion environme.nt, falls under the incremental change~. Throughput, defi ned as the rat io of total sat isfied demand 10
the tota l generated demand (per cent of customer orders fulfilled), could be considered a good indicator of a stabilised manufacturing system. A re lated incliceHar, somewhat dependent on the technology in operat ion, and hence whose use shou ld remain limited to the benchmarking, is the ratio of WIP inventory levels at \ arious celis/workstati ons to Ihe throughput ratc of the sys tem, under th rou ghput max im isa ti on. As
7
2
2
6
4
4
4
(per ce nt) (x I OO Units)
8 9 10
2 0 0 99.78 18
4 2 2 99 .<)0 30
0 0 6 99 .90 28
6 () () 57.62 2()
4 6 2 52 .37 26
4 2 0 57 .62 IS
cla imed in thi s paper, it was the distributi on of W1P along the cells and not the total WIP, that was the mos l il11p~ rtan~ process indica tor of throug hput maXlmtsatlon. Continuous updati on of databases can be incorporated so as to accommodate as many customer orders as possib le into the producti on sc hed ule. Li ne balanci ng studies (in preceding and succeeding departments) can be conducted for uniform cycle timc (minimum idle time) in the production syste m. GPSS model parameters are req uired to be updated based on the actual de mand . Model can be updated by co nsic1-e.ring variable setup time, breakdown time and rejecti on rate, etc. Altern,lte simu lation model. can be developed and compared and suitable model ean be selected.
Acknowledgements
The Authors wOllld like to thank the C hairman, and G Ravi chandran, Managcr, Manufacturing Systems
--f~
VENKATARAM ANA IAH el 01. : JlT ENVI RONMENT UNDER CELLULAR LAYOUT 263
Redesign ofTl Cycles Ltd. Our thanks are also due to Professor TT Narendran and anonymous referee for their valuable suggestions .
4 Mejabi , 0 & Wasserman S G, Simulation constructs fo r JIT modelling, Inte J Prod Res, 30(5), ( 1992 ). 1119-
1135.
References
Berkley B J, Testing min im um performance level of kanban controlled 1i 1 • .;:S, Int J Prod Res, 32 ( I ), ( 1994)
93- 109. 2 Burbidge J L, Change to group technology: process
organisat ion is obsolete, IlIfer J Pmd Res , 30 (5), ( 1992)
1209-1219. 3 Krajewsk i L J & Ritzman L P, Operations lIlanage
ment-strategy and analvsis, (Addision Wes ley Inc.) ( 1996)
5 Schroer B J, Black J T and Zhangc S X, JIT. with kanban, manufacturing system simulation on a microcolllputer, Silllulation COlIl/cil, II/ e. ( 1985) 62-70.
6 Sarker B R, Simulati on of ajust in timc producti on systcm, COIllP Ind 16, ( 1989), 127-137.
7 Schriber T J, Sim.ulation using C PSS, (Joh ll Wi ley &
sons) ( 1974).
8 Yavuz,H and Satir, A Kanban based simulati on study ofa mixed model j ust in lime manufactu ri ng line, Inter J Pmd Res, 33(4), ( 1995) 1027-1 04~ .
About the authors S Vellkataramanaiah, Cllrrentir working fo r his Doctoral deg ree in th e area of Ceffltlar MWIl t/acwring. Cu rrenti" he is on (/ one year assignlllent at Indion Institllte of Management. Bangalore I\ 'orking on a sponsored Research Project en titled "Ma llagerialfram.e workfor illlplementation ofCellltlar Manufacturing - In Indian Con text ". He has presented several technica l papers at national and in ter
national level con{erences in the area o!' POM. Indll st ria l Engineering Division. Departlll en t o!,Mon
agelll ent Studies. College of Eng ineering, Cllindy. Anna Vni l'e rsity, Chen/w i, India- 600025
K Krishllaiah is Head of the Industrial Engineering. who fwd abollt 6 years of industrial experience and J 9 years of teaching experience. His articfes appeared in narional and internationaljollma/s and Inesented several papers at national and intemational cor~j'erences. His arl!([S of interest in elude TQM,
work design/e rgonolll ics. Operations Managelllent & CIMS. Indllstriol Engineering Division. College
of Eng illecrill /? Cllillriy. All/ I({ Vlli\'ers it.\'. Chellllai, India 600025
R.Pichandi is A llditor SofMare quality ill th e ({reo o{ Y2K implemelltat ion as Asst. SYst em Ellgill ce /; 560 052