Assembly Line Balancing

The assembly line is a production line

where material moves continuously
through a series of workstations where
assembly work is performed.

1
 Assembly Lines
• Principle of Interchangeability
– individual components that make up a finished
product should be interchangeable between
product units
• Division of Labor – complex activities
divided into elemental tasks
– work simplification
– standardization
– specialization
• Mass Production

2
 Production Systems

 Project Shop project networks

 Job Shop job shop scheduling
 Flow Shop flow shop scheduling
 Assembly Line assembly line balancing
– e.g. cyclic scheduling
 Continuous Flow single facility EOQ
model

3
 The Problem

Assign work elements (tasks) to workstations

to minimize unit assembly costs (e.g. labor cost).

flow of the line

station 1 station 2 station 3

Tasks 1 2 3 4 5 6
precedence precedence
requirements requirements
4
 Cycle Time
The time between the completion of two successive
products, assumed constant for all products for a
given production line speed. The minimum value
of the cycle time must be greater than or equal to
the longest station time.

A group of engineering management

students discussing cycle times.

5
 Problem Formulation
Assume production rate of P with m parallel lines.
Then each line must produce a unit every m / P time units.

Set Cycle time = C <= m / P ; the time between completion of

two successive units.

Example:
Planned order release requires a production rate of 80
units per hours. Four (4) assembly lines are available.
Therefore cycle time = C  4/80 = .05 hr. = 3 minutes

6
 Station Time

Let ti = time to perform task i where i = 1,2,…,n

Sj = station j time where

S j   ti Sj <= C
iI j

and Ij = {i | task i is assigned to station j}

7
 Performance Measures
let k = number of workstations; 1 <= k <= n

dj = C – Sj = delay (idle) time at station j

k k n
total idle time: IT  �d  � C  S   k C  �t
j 1
j
j 1
j
i 1
i

S j 1
j
line efficiency: LE  x 100% 0 is perfect
kC balance

 S  Sj 
2
line smoothness index: SI  max
j 1
8
 Minimizing Idle Time
minimizes assembly time per unit
n
Min IT  k C   ti
i 1 for IT = 0,
n
must be integer

t i
s.t. k  i 1
S j   ti  C
C iI j
kn
 n m
tmax  C  min  ti , 
 i 1 P 
This looks like
precedence requirements it is NP-hard to me.
9
 Precedence Relationships

• Precedence constraints
– some tasks may have to be completed in a
particular sequence task i task j
• Zoning restrictions
– some tasks cannot be performed at the same
workstation (divorces)
– some tasks may be required to be performed at
the same workstation (marriages)

10
 Our Very First Example
S1 S2 S3 S4 S5

1 2 3 4 5
5 min 7 min 10 min 6 min 8 min

n
k = 5 C = 10 min. P = 6 per hr. t
i 1
i  36

Performance Measures
5

� 10  S 
IT = 5(10) – 36 = 14 min. 2
LE = 36/50 = 72% SI  j
j 1
SI = 7.35 11
 Our Very First Example
S1 S2 S3 S4

1 2 3 4 5
5 min 7 min 10 min 6 min 8 min

n
k = 4 C = 12 min. P = 5 per hr. t
i 1
i  36

Performance Measures
IT = 4(12) – 36 = 12 min.
LE = 36/48 = 75%
SI = 7.48
12
 Our Very First Example
S1 S3
S2

1 2 3 4 5
5 min 7 min 10 min 6 min 8 min

n
k = 3 C = 14 min. P = 4.286 per hr. t
i 1
i  36

Performance Measures
IT = 3(14) – 36 = 6 min.
LE = 36/42 = 85.7 %
SI = 4.47
13
 Our Very First Example
S1 S2

1 2 3 4 5
5 min 7 min 10 min 6 min 8 min

n
k = 2 C = 22 min. P = 2.72 per hr. t
i 1
i  36

Performance Measures
IT = 2(22) – 36 = 8 min.
LE = 36/44 = 81.8 %
SI = 8
14
 Our Very First Example
S1

1 2 3 4 5
5 min 7 min 10 min 6 min 8 min

n
k = 1 C = 36 min. P = 1.67 per hr. t
i 1
i  36
Look, a
Performance Measures perfectly
IT = 1(36) – 36 = 0 min. balanced line.
LE = 36/36 = 100 %
SI = 0
15
 Chuck. Could you
summarize all this for me?
Just tell me what I need to
know!

k C P IT LE SI
5 10 min 6/hr 15 min 72 % 7.35
4 12 min 5/hr 12 min 75 % 7.48
3 14 min 4.28/hr 6 min 85.7 % 4.47
2 22 min 2.73/hr 8 min 81.8 % 8
1 36 min 1.67/hr 0 100 % 0

16
 Our Very Next Example Problem
Task 1 2 3 4 5 6 7
ti 5 3 4 3 6 5 2

Task 8 9 10 11 12
ti 6 1 4 4 7

7 8
2 3

1 6 9 12
4 5
10 11
precedence relationships 17
 Trial and Error Approach
• Find minimum number of stations for a given cycle
time
• Repeat for various cycle times
• Select solution that minimizes idle time
12

t
i 1
i  50  2 x 5 x 5 prime
factors
n
cycle times feasible? min k Min IT  k C  �ti
50 yes 1 i 1

25 yes 2 n

10 yes 5 �t i
Then k  i 1
an integer
5 no 10 tmax = 7 C
2 no 25 18
 I II III IV V VI VII

2 3 7 8

1 6 9 12
4 5
10 11
station task ti column sum cumulative sum
I 1 5 5 5 C = 10
II 2 3
4 3 6 11 IT = 7(10) – 50 = 20
III 3 4
5 6 10 21 LE = 50/70 = 71%
IV 6 5 5 26
V 7 2 SI = 9.16
9 1
10 4 7 33
VI 8 6
11 4 10 43
19
VII 12 7 7 50
 Heuristic
• place each task as far to the left as possible
• no restriction of movement within a column
• can move tasks further to the right
• assign tasks to work stations such that the
sum of the times does not exceed C
• always select task with longest time when
forming a workstation

20
I II III IV V VI
task ti pred
1 5 0
2 3 7 8 2 3 1
3 4 2
4 3 1
1 6 9 12 5 6 2
6 5 5
4 5
X 10 11
7 2 6
8 6 7
9 1 6
station task ti column sum cumulative sum
10 4 6
I 1 5
11 4 7
2 3 8 8
II 4 3 12 7 11
5 6 9 17
III 3 4
6 5 9 26 C = 10
IV 7 2
9 1 IT = 6(10) – 50 = 10
10 4 7 33
V 8 6 LE = 50/60 = 83%
11 4 10 43
VI 12 7 7 50 SI = 4.89 21
 I II III IV V VI
Can move
1 4 3 7 10 9 tasks to the
right.
2 5 6 8 11 12

station task ti column sum cumulative sum

I 1 5 C=9
2 3 8 8
II 4 3 IT = 6(9) – 50 = 4
5 6 9 17
III 3 4 LE = 50/54 = 92.6%
6 5 9 26
IV 7 2 SI = 2
8 6 8 34
V 10 4
11 4 8 42
VI 9 1
22
12 7 8 50
 Positional Weight Method
• Find positional Weight (PW) for each task
• Rank tasks based upon PW
– highest first
• Assign tasks to stations with highest rank first
• Continue to assign tasks as long as time remains
– task does not violate precedence relationship
– station time does not exceed cycle time
• Repeat until all tasks are assigned
• Each task is assigned to the first feasible station
– greedy algorithm

23
 Positional Weight
PWi = time of the longest path from beginning of task i
through the remainder of network.
3 4 2 6
2 3 7 8
5 5 7
1
1 3 6 6 9 12
4 5 4 4
10 11

Task 1 2 3 4 5 6 7
PWi 34 27 24 29 26 20 15

Task 8 9 10 11 12
PWi 13 8 15 11 7 24
Task 1 2 3 4 5 6 7
PWi 34 27 24 29 26 20 15

Task 8 9 10 11 12
PWi 13 8 15 11 7

Rank Task PW task ti pred

1 1 34 1 5 0
Assume CT = 10 2 3 1
2 4 29
3 4 2
3 2 27 4 3 1
4 5 26 5 6 2
5 3 24 6 5 5
6 6 20 7 2 6
7 7 15 8 6 7
8 10 15 9 1 6
9 8 13 10 4 6
10 11 11 11 4 7
11 9 8 12 7 11
12 12 7 25
ank Task PW station task ti column sum cumulative task ti pred
1 1 34 I 1 5 1 5 0
2 4 29 4 3 8 8 2 3 1
3 2 27 II 2 3 3 4 2
4 5 25 5 6 9 17 4 3 1
5 3 24 III 3 4 5 6 2
6 6 20 6 5 9 26 6 5 5
7 7 15 IV 7 2 7 2 6
8 10 15 10 4 6 32 8 6 7
9 8 13 V 8 6 9 1 6
10 11 11 11 4 10 42 10 4 6
11 9 8 VI 9 1 11 4 7
12 12 7 12 7 8 50 12 7 11

C = 10 I have long advocated the

positional weight method
in order to achieve the best
IT = 6(10) – 50 = 10
balance.
LE = 50/60 = 83.3%
SI = 5.09 26
An Integer Programming Approach
let xik = 1 if task i assigned to station k; 0 otherwise
ci,k = cost coefficient of xi,k where cik < ci,k+1
n K
Min  c
i 1 k 1
x
ik ik

n
s.t. t x
i 1
i ik C k  1,..., K
K

x
k 1
ik 1 i  1,..., n
h
xih   x jk precedence relationship
k 1 when task j precedes task i
27
 Some Final Considerations
• If significant idle time remains
– consider parallel lines with larger cycle times
– use more than one worker per station (group stations)
• task variability
– max station time = E[Si] + 2.33 STD[Si] <= C
– probability all stations complete on time .99k
• provide rework area These are
some very
• add buffers good final
• use unpaced (asynchronous) lines considerations.
• Max profit rather then minimize idle time

28