An Overview of Design and Operational Issues of Kanban Systems

1
An Overview of Design and Operational Issues of
Kanban Systems

• by M. S. Aktürk and F. Erhun

Presented by Ali Koç 16 December 2009
2
Outline

• JIT approach and Kanban Systems
• Overview of Literature
• Computational Analysis
• Results and Implications

3
JIT

• Having necessary amount of material available
  where is needed and when is needed.
• Effective when high process reliability, low
  set-up times, and low demand variability
• Pull system of production
• More successful in repetitive manufacturing
  environment
• Large information lead-times

4
Advantages of JIT

• Reduced inventories
• Reduced lead-times
• Higher quality
• Reduced scraprework rates
• Increased flexibility
• Easier automation
• Better utilization

5
Kanban

• Shop floor control of JIT philosophy
• Japanese word for visual record or card
• Manual method of harmoniously controlling
  production and inventory quantities
• Kanbans used for, controlling in-process
  inventory as well as pulling the parts

6
Types of Kanban Systems

• Dual-card kanban systems
• Production Kanban defines the quantity of
  specific parts to be produced
• Transportation Kanban defines the quantity of
  parts to be withdrawed from the succeeding
  station

• Single-card kanban systems
• For MH, still transportation kanban, whereas for
  production, schedule is provided by the central
  production planning

7
Kanban Systems

• Single-card kanban systems
• More similar to the push systems
• Easier to implement
• Shorter information lead-time
• Can also be used when the succeeding stage is
  closed

• In this study single-card kanban system is used

8
Review Period of Kanban Systems

• Instantaneous Review Systems
• Periodic Review Systems
• Fixed quantity, non-constant withdrawal cycle,
  waits for a predetermined quantity to dispatch
• Fixed withdrawal cycle, non-constant quantity,
  waits for a fixed time unit to dispatch

9
Literature Review

• Determining the design parameters

• Determining the design parameters

10
Literature Review
STUDIES
11
Literature Review
STUDIES
12
Literature (Findings)

• Most of the papers consider kanban size and the
  number of kanbans separately.
• Variability in processing times and demand rates
  are amplified in a multi-stage setting. Excess
  capacity should be available
• Number of kanbans is a key factor for
  performance.
• For smooth operation, the stages should be
  balanced and suppliers should be reliable. As
  demand variability increases system performance
  decreases
• Kanban size has significant effect on system
  performance. Smaller kanban sizes cause smaller
  in-process inventory and better customer service

13
Literature Review

• Determining the kanban sequences

• FCFS, SPT, SPT/LATE, HPF, HPF/LATE, and FCFS/SPT
  dispatching rules are studied in the literature
• Sequencing in kanban-controlled shops are more
  complex then conventional sequencing problem
• More sophisticated scheduling rules should be
  used to determine the effects of scheduling on
  performance of kanban systems

14
Computational Analysis
Layout
15
Computational Analysis

• Testing the design parameters under some
  operating parameters
• Operating parameters (4 dispatching rules)
• FCFS
• FCFS-F
• SPT / LATE
• SPT-F

16
Operating Parameters

• FCFS, assigns the maximum waiting item

• SPT / LATE,
• If, maximum absolute delay is higher then a
  predetermined level uses FCFS
• Else, assigns the minimum updated processing
  times, if tie, use FCFS

17
Operating Parameters

• FCFS-F, assigns the maximum backlogged family, if
  tie, use SPT-F

• SPT-F, selects the family with minimum processing
  time, if tie, use FCFS-F

18
Computational Analysis

• Design Parameters (Experimental Factors)

19
Design Parameters

• Demand Distribution

• Backorder/Inventory Ratio
• bijm(B/I)hijm

20
Design Parameters

• Balance of Load
• In balanced case, the processing times of each
  item at each stage is selected from UN
  0.1,0.3
• In unbalanced case, only stage D has different
  processing time
• When number of parts in each family is low,
  processing time of stage D is UN 0.3,0.5 and
  when high, from UN 0.2,0.4

21
Design Parameters

• Setup Time
• Follows UN SLm, SHm
• SLm(S/P)x(average processing time of family i at
  stage m)x0.5
• SLm(S/P)x(average processing time of family i at
  stage m)x1.5

22
Other Parameters

• Inventory holding cost (hijm) is generated from
  UN 5,10
• Max. inv. level na DL(1s)
• Lead-time depends on work-in-queue
  (RagatzMabert)
• Varying max. inv. level, depending on queue,
  provides flexibility, increase solution space

23
Other Parameters

• Six alternative withdrawal cycle length
  480,240,120,60,30,15 minutes

• Vary number of kanbans and kanban sizes in
  coordination to keep MINVijm constant
  (KarmarkarKekre).

• A total of 36 alternatives for each of 27
  experiment

24
Results and Implications

• For family-based methods (S/P) ratio is smaller,
  forcing shorter withdrawal cycle lengths

25
Results and Implications

• Positive correlation between withdrawal cycle
  length and maximum inventory level (FCFS is
  highest)
• Max. inv. level is not alone a performance
  measure. When interpret with the cost, the best
  performance is by SPT/LATE (55.88)

26
Results and Implications

• SPT/LATE is the best again in backorder cost and
  fill rate
• So, more sophisticated rules are needed for
  better performance

27
Results and Implications

• 640 runs taken to test the significance of the
  results
• For inv. hold. cost, pairs are different for
  plt0.000 except FCFS-F and SPT/LATE pair for which
  plt0.002
• For backorder cost, pairs are different for
  plt0.000 except FCFS-F and SPT-F pair for which
  plt0.005

• Relative success of SPT/LATE is attributed to the
  fact that it is a composite dispatching rule,
  combining family and item-based methods
• Minimum cost not necessarily implies minimum T

28
Experimental Analysis on Results

• For each sequencing rule, two-way ANOVA is
  applied.
• For inv. hol. cost, all of the factors are
  statistically significant, with plt0.00
• For backorder cost, all of the factors except S/P
  for all rules and imbalance for FCFS are
  statistically significant, with plt0.00

29
Results and Implications

• Factors A,B, and C affect the inv. hold. and
  backorder cost, since they affect the amount to
  produce
• Imbalance always decreases systems performance
• FCFS is not affected by imbalance, since it is
  applies consistent priorities
• As S/P increase, inventory level, inventory
  holding cost increase, and fill rate decreases,
  since setup increases batch size, and in turn,
  cycle length
• Demand variability decreases fill rates, since it
  inflates lead-times

30
Last Word

• Very few studies exist that consider the kanban
  sizes explicitly. In fact, the number of kanbans
  required depends on kanban sizes, therefore these
  parameters should be determined simultaneously,
  not sequentially.
• The existing JIT production planning models deal
  mainly with smoothing production schedules. None
  of the studies consider the impact of operational
  issues on the design parameters.
• Even though dual-card kanban systems are periodic
  in nature, there are a limited number of studies
  on periodic review systems.

31

• Q/A