15/04/2024
Lean Thinking & Lean Tools
- Flow without Interruption: Session 2
Professor Dongping Song
Liverpool University Management School
Email: [email protected]
1
Learning Outcomes
• Techniques to Solve Line Balancing
• Longest-Task-Time (LTT) Heuristic; STT
• Kilbridge and Wester Method (KWM)
• Ranked Positional Weights (RPW) Method
• Incremental Utilisation Heuristic
• Other Ways To Improve Line Balance Computer Method
Mathematical Methods
• Mixed-model production leveling
• Weighted Average Time method/rule
• Production levelling in logistics
2
1
15/04/2024
Review Terminologies
• Task –the elements of work; operation
• Task precedence
• Task time (also operation cycle time)
• Workstation
• Work center
• Production rate
• Cycle time (line cycle time)
• Stacked time
• Takt time
3
Techniques to Solve Line Balancing &
Production Levelling
• Heuristic Methods -- based on simple rules:
– Longest-Task-Time Heuristic
– Shortest-Task-Time Heuristic
– KWM --Kilbridgeand WesterMethod (1961)
– Ranked Positional Weights Method
– Incremental Utilisation Heuristic
• Computerised Tools, e.g. COMSOAL (1996)
• Mathematical Methods
– Linear/Integer Programming
– Dynamic Programming
4
2
15/04/2024
Line Balancing - Longest-Task-Time
Heuristic
• LTTis also called Largest-Candidate Rule (LCR).
• Step 1. Construct the precedence diagram. List all elements
(tasks) in descending order of Te(= task time) value.
• Step 2. To assign elements to the first workstation, start at the
top of the list for the available elements, selecting the first
feasible elementfor placement at the station. A feasible
elementis one that satisfies the precedence requirements
and does not cause the sum of the Tevalue at workstation to
exceed the cycle time CT. Open a new workstation as needed.
• Step 3. Repeat step 2.
Select the available and feasible task with the largest Te.
5
Line Balancing – Kilbridge and Wester
Method
• Step 1. Construct the precedence diagram so those nodes
representing work elements of identical precedence are
arranged verticallyin columns.
• Step 2. List the elements in order of their columns, column I
at the top of the list. If an element can be located in more
than one column, list all columns by the element to show the
transferability of the element.
• Step 3. To assign elements to workstations, start with the
column I elements. Continue the assignment procedure in
order of column number until the cycle time is reached (CT).
Select tasks according to their position in the precedence diagram
6
3
15/04/2024
Line Balancing – Ranked Positional
Weights Method
• Introduced by Helgeson and Birnie in 1961.
• Combined the LTT and KWM methods.
• The RPW takes account of both the Te value of the
element and its position in the precedence diagram.
• Need to calculate RPW for all elements;
• Then, the elements are assigned to workstations in
the general order of their RPW values.
7
Procedure of RPW
• Step 1. Construct the precedence diagram. Then calculate the
RPW for each element by summing the elements Tetogether
with the Tevalues for all the elements that follow it in the
arrow chain of the precedence diagram.
• Step 2. List the elements in the order of their RPW, largest
RPW at the top of the list. For convenience, include the Te
value and immediate predecessors for each element.
• Step 3. Assign elements to stations according to RPW priority
rule, avoiding precedence constraint and cycle time violations.
Select tasks considering both their position and time value.
8
4
15/04/2024
Example: Small electrical appliance
No. Task description Time (min) Precedence by
1 Place frame on work holder and clamp 0.2 --
2 Assemble plug, grommet to power cord 0.4 --
3 Assemblebrackets to frame 0.7 1
4 Wire power cord to motor 0.1 1,2
5 Wire power cord to switch 0.3 2
6 Assemblemechanism plate to bracket 0.11 3
7 Assemble blade to bracket 0.32 3
8 Assemble motor to bracket 0.6 3,4
9 Align blade and attach to motor 0.27 6,7,8
10 Assemble switchto motor bracket 0.38 5,8
11 Attach cover, inspect, and text 0.5 9,10
12 Place in tote pan for packing 0.12 11
Assume cycle time is 1 minute. To determine number of WS and assignment of tasks.
9
Example: Precedence Diagram
0.11
6
0.7
0.32 0.27
3
0.2
7 9
1
0.1 0.5 0.12
4 11 12
0.6 0.38
0.4
8 10
2
0.3
5
10
5
15/04/2024
Example: LTT – Step 1
Task Time Immediate
(Te) predecessor
3 0.7 1
8 0.6 3,4
11 0.5 9,10
2 0.4 ---
10 0.38 5,8
7 0.32 3
5 0.3 2
9 0.27 6,7,8
1 0.2 ---
12 0.12 11
6 0.11 3
4 0.1 1,2
List all tasks in descending order of task time
11
Example: LTT – Step 2,3
Task Time Immediate
WS Task Te sum(Te) Available
(Te) predecessor
3 0.7 1 1 2 0.4 (2, 1)
8 0.6 3,4 5 0.3 (5, 1)
11 0.5 9,10 1 0.2 (1)
2 0.4 --- 4 0.1 1.00 (3, 4)
10 0.38 5,8 2 3 0.7 (3)
7 0.32 3 6 0.11 0.81 (8, 7, 6)
5 0.3 2 3 8 0.6 (8, 7)
9 0.27 6,7,8 10 0.38 0.98 (7, 10)
1 0.2 --- 4 7 0.32 (7)
12 0.12 11 9 0.27 0.59 (9)
6 0.11 3 5 11 0.5 (11)
4 0.1 1,2 12 0.11 0.62 (12)
6
3
7 9
1
1 1
4
1 1 2
2 8
0
5
12
6
15/04/2024
Example: KWM – Step 1
I II III IV V VI
0.11
6
0.7
0.32 0.27
3
0.2
7 9
1
0.1 0.5 0.12
4 11 12
0.6 0.38
0.4
8 10
2
0.3
5
Arrange the tasks in columns vertically
13
Example: KWM – Step 2
Task Column Time Sum of
(Te) Column Te
1 I 0.2
2 I 0.4 0.6
3 II 0.7
4 II 0.1
5 II,III 0.3 1.1
6 III 0.11
7 III 0.32
8 III 0.6 1.03
9 IV 0.27
10 IV 0.38 0.65
11 V 0.5
12 VI 0.12 0.62
List the tasks in order of their columns
14
7
15/04/2024
Example: KWM – Step 3
Task Column Time Sum of
WS Task Te sum(Te)
(Te) Column Te
1 I 0.2 1 1 0.2
2 I 0.4 0.6 2 0.4
3 II 0.7 4 0.1
4 II 0.1 5 0.3 1.00
5 II,III 0.3 1.1 2 3 0.7
6 III 0.11 6 0.11 0.81
7 III 0.32 3 7 0.32
8 III 0.6 1.03 8 0.6 0.92
9 IV 0.27 4 9 0.27
10 IV 0.38 0.65 10 0.38 0.65
11 V 0.5 5 11 0.5
12 VI 0.12 0.62 12 0.11 0.62
Assign the tasks into workstations
15
Example: RPW – Step 1
• For Task 1, the tasks that follow it in the arrow chain are
3,4,6,7,8,9,10,11,12.
• The Ranked Positional Weight (RPW) for Task 1 would be the sum
of the Tefor all these tasks, plus Tefor Task 1, which is 3.3.
• The same can be done for other Tasks.
0.11 Calculate RPWs for all tasks
6
0.7
3 0.32 0.27
0.2
7 9
1
0.1 0.5 0.12
4 11 12
0.6 0.38
0.4
8 10
2
0.3
5
16
8
15/04/2024
Example: RPW – Step 2
Task RPW Te Immediate
predecessor
1 3.3 0.2 ---
3 3 0.7 1
2 2.67 0.4 ---
4 1.97 0.1 1,2
8 1.87 0.6 3,4
5 1.3 0.3 2
7 1.21 0.32 3
6 1.00 0.11 3
10 1.00 0.38 5,8
9 0.89 0.27 6,7,8
11 0.62 0.5 9,10
12 0.12 0.12 11
List all tasks in descending order of RPW
17
Example: RPW – Step 3
Task RPW Te Immediate
WS Task Te sum(Te) Available
predecessor
1 3.3 0.2 --- 1 1 0.2 (2, 1)
3 3 0.7 1 3 0.7 0.90 (2, 3)
2 2.67 0.4 --- 2 2 0.4 (2,6,7)
4 1.97 0.1 1,2 4 0.1 (4,5,6,7)
8 1.87 0.6 3,4 5 0.3 (5,6,7,8)
5 1.3 0.3 2 6 0.11 0.91 (6,7,8)
7 1.21 0.32 3 3 8 0.6 (7,8)
6 1.00 0.11 3 7 0.32 0.92 (7,10)
10 1.00 0.38 5,8 4 10 0.38 (10,9)
9 0.89 0.27 6,7,8 9 0.27 0.65 (9)
11 0.62 0.5 9,10 5 11 0.5 (11)
12 0.12 0.12 11 12 0.11 0.62 (12)
6
3
7 9
1
1 1
4
1 1 2
2 8
0
5
18
9
15/04/2024
Example: STT
Task Time Immediate
(Te) predecessor WS Task Te sum(Te) Available
3 0.7 1 1 1 0.2 (2, 1)
8 0.6 3,4 2 0.4 (2, 3)
11 0.5 9,10 4 0.1 (3,4,5)
2 0.4 --- 5 0.3 1.00 (3,5)
10 0.38 5,8 2 3 0.7 (3)
7 0.32 3 6 0.11 0.81 (6,7,8)
5 0.3 2 3 7 0.32 (7,8)
9 0.27 6,7,8 8 0.6 0.92 (8)
1 0.2 --- 4 9 0.27 (9,10)
12 0.12 11 10 0.38 0.65 (10)
6 0.11 3 5 11 0.5 (11)
4 0.1 1,2 12 0.11 0.62 (12)
6
3
7 9
1
1 1
4
1 1 2
2 8
0
5
19
Example – LTT, KWM, RPW, STT
WS LTT KWM RPW STT
1 1.00 1.00 0.90 1.00
2 0.81 0.81 0.91 0.81
3
0.59 0.92 0.92 0.92
4 0.98 0.65 0.65 0.65
5
0.62 0.62 0.62 0.62
20
10
15/04/2024
Result Analysis (RPW)
Line Balance Ratio = Stacked Times . X 100%
Largest Op. x No. of Operators
Line Balance Ratio = 4 . X 100%
0.92 x 5
= 87 %
21
Result Analysis (RPW)
Line Balance Eff = Stacked Times . X 100%
TaktTime. x No. of Operators
Assume Takt time = CT
Line Balance Eff. = 4 . X 100%
1 x 5
= 80 %
22
11
15/04/2024
Result Analysis (LTT vs. RPW)
Rules LTT RPW
Takttime 1 1
No of WSs 5 5
Maximum WS time 1 0.92
Balance loss 1 1
Line balance ratio 80% 87%
Line balance efficiency 80% 80%
23
Learning Outcomes
• Longest-Task-time Heuristic
• Kilbridgeand WesterMethod
• Ranked Positional Weights Method
• Compare LTT, STT, KWM, and RPW
• Incremental Utilisation Heuristic
• Other Ways To Improve Line Balance
• Product leveling
• Mixed-model production leveling
• Weighted Average Time method/rule
• Remarks on production levelling
24
12
15/04/2024
Line Balancing – Incremental
Utilization Heuristic
• What if task times are greater than the cycle time?
• Assign tasks to a number of workstations (or
operators) to form a number of work centres
• Using Incremental Utilization Heuristic
– Tasks are added to each WS in order of task precedence
one at a time until utilization is 100% or falling.
– To determine: number of work centres, number of
workstations at each work centre, assignment of tasks to
each work centre.
Utilisation = sum of tasks assigned / (CT * NoOfWS)
25
Example
Textech, a large electronics manufacturer, assembles model AT7S handheld
calculators at its Midland , Texas, plant. The assembly tasks that must be
performed on each calculator are shown in the Table below.
• The parts used in this assembly line are supplied by materials-handling
personnel to parts bins used in each task. The assemblies are moved along
by belt conveyors between workstations. The productive time is 54
minutes per hour.
• Textechwants this assembly line to produce 540 calculators per hour.
Assuming 100% production line efficiency.
1. Compute the cycle time per calculator in minutes.
2. Compute the minimum number of workstations.
3. How would you combine the tasks into workstations and work
centres to minimize idle time? Evaluate your proposal.
26
13
15/04/2024
Task Description Immed. Preced. Task time
A place circuit frame on jig. none 0.18
B place circuit #1 into frame. A 0.12
C place circuit #2 into frame. A 0.32
D place circuit #3 into frame. A 0.45
E Attach circuits to frame. B,C,D 0.51
F Solder circuit connections to central circuit control. E 0.55
G Place circuit assembly in calculator inner frame. F 0.38
H Attach circuit assembly to calculator inner frame. G 0.42
I Place and attach display to inner frame. H 0.30
J Place an attach keyboard to inner frame. I 0.18
K Place and atach top body of calculator to inner frame. J 0.36
L Place and attach power assembly to inner frame. J 0.42
M Place and attach bottom assembly to inner frame. K,L 0.48
N Test circuit integrity. M 0.30
O Place calculator and printed matter in box. N 0.39
27
Precedence Diagram
B K
/ \ / \
A – C – E – F – G – H – I – J M – N – O
\ / \ /
D L
28
14
15/04/2024
Example
Compute the target cycle time per unit:
A=.18
Takt Time= productive time per hour/demand per hour B=.12
C=.32
= 54/540 = 0.10 minute
D=.45
Cycle Time = TT= 0.10 minute E=.51
F=.55
G=.38
Compute the minimum number of WS: H=.42
I=.30
minNumWS= Sum of task times / CT J=.18
K=.36
= 5.36 / 0.10
L=.42
= 53.60 WS M=.48
N=.30
:= 54 WS
O=.39
29
A=.18
B=.12
C=.32
D=.45
E=.51
F=.55
G=.38
H=.42
I=.30
J=.18
B K K=.36
/ \ / \ L=.42
A – C – E – F – G – H – I – J M – N – O M=.48
N=.30
\ / \ /
O=.39
D L
CT = 0.1; Utilisation = sum of tasks assigned / (CT * NoOfWS);
30
15
15/04/2024
B K
/ \ / \ A=.18
A – C – E – F – G – H – I – J M – N – O B=.12
\ / \ / C=.32
D L D=.45
E=.51
F=.55
G=.38
H=.42
I=.30
J=.18
K=.36
L=.42
M=.48
N=.30
O=.39
CT = 0.1; Utilisation = sum of tasks assigned / (CT * NoOfWS);
31
Summarize the assignment of tasks to
WS
Tasks in work A,B C,D,E F,G,H,J J,K,L,M,O
centers
Work centers 1 2 3 4
Actual no of WS 3 13 17 22
Total number of WS = 3+13+17+22 = 55.
Average Utilisation = minNumOfWS / actualNumOfWS
= 53.6/55
= 97.5%
32
16
15/04/2024
Computerized Line Balancing
• Line balancing by hand becomes unwieldy as the problems
grow in size.
• There are software packages that will balance large lines
quickly. IBM'sCOMSOAL(Computer Method for Sequencing
Operations for Assembly Lines) and GE's ASYBL (Assembly
Line Configuration Program) can assign hundreds of work
elements to workstations on an assembly line.
• Random trial and search for better solutions.
• Some software lets the user select specific heuristic, e.g.
ranked positional weight (RPW), longest operation time (LTT),
shortest operation time (STT), most number of following
tasks, and least number of following tasks.
33
COMSOAL: Computer Method
Operation sequences are generated by randomly picking a task and
constructing subsequent tasks
New stations are opened when needed
New solution that exceeds the best solution so far is discarded
Better solutions become upper bounds
• NIP(i) = Set of Immediate Predecessors for each task i;
• WIP(i) = Set of tasks for which i is an immediate predecessor
• TK = Set of N tasks
• A = Set of unassigned tasks
• B = Set of tasks from A with all immediate predecessors assigned
• F = Set of tasks from B with tasks times not exceeding remaining cycle
time in the current workstation
34
17
15/04/2024
COMSOAL: Algorithm
For generating X trial solutions to minimise the total idle time:
1. Initialisation of algorithm
1.-SET x=0, UB=inf, c=CT 2. Initialisation at each iteration
2.-START NEW SOLUTION: SET x=x+1, A=TK, NIPW(i) = NIP(i)
3.-PRECEDENCE FEASIBILITY: FOR i IN A, IF NIPW(i) = 0 , ADD i TO B
4.-TIME FEASIBILITY: 3. Update available tasks in B
FOR i IN B, IF ti< c ADD i TO F.
4. Update feasible tasks in F
If F empty , go to 5 , otherwise go to 6
5.-OPEN NEW STATION:
IDLE=IDLE + c , c = CT.
5. Discard a bad solution
If IDLE > UB , go to 2, otherwise go to 3
A = Set of unassigned tasks;
B = Set of available tasks from A;
F = Set of feasible tasks from B; c = the remaining capacity of a WS
35
COMSOAL: Algorithm
6.-SELECT TASK: SET m = card{F} = size of set F;
6. Randomly select tasks in F
RANDOM GENERATE RN in U(0,1);
LET i* = [m*RN]thTASK from F;
6. Task i* has been assigned,
REMOVE i* from A, B, F;
update A, B, F.
c = c –ti;
FOR ALL j in WIP(i*), NIPW(j) = NIPW(j) –{i*};
IF A EMPTY --> 7, OTHERWISE --> 3
6. Update preceding tasks
7.-SCHEDULE COMPLETION
IDLE = IDLE + c
IF IDLE < UB , UB = IDLE --> STORE SCHEDULE
IF x = X , STOP, OTHERWISE --> 2 7. Output the solution of
the assigned tasks
A = Set of unassigned tasks;
B = Set of available tasks from A;
F = Set of feasible tasks from B; c = the remaining capacity of a WS
36
18
15/04/2024
Mathematical Methods (IP)
Notation:
• t = task time;
j
• NIP = set of Immediate Predecessors for task j;
j
Decisions:
• F n= 1 if WS n is used/opened; 0 otherwise. • Number of WSs;
• X = 1 if task j is assigned to WS n; 0 otherwise. • Assignment of tasks;
jn
Objective: MIN F minimum number of WS to be used.
n
1. Each task has to be assigned to one and
Constraints:
only one WS;
1. X = 1, for any j; 2. Tasks on a WS should not exceed CT;
n jn
2. X * t <= CT*F , for any n; 3. Lower number WSs fill up first;
j jn j n
4. Precedence relationship constraints.
3. F >= F , for any n;
n n+1
4. X <= (X + X + … + X ),for any j, n; and any k in NIP;
jn k1 k2 kn j
37
More Approaches To Improve Line
Balance
• Dividing work elements
• Changing speed/feed at stations
• Activity/process analysis
– Better workplace layout, redesign tooling & fixturing
• Pre-assembly of components
• Merge tasks/workers
• Parallel stations
38
19
15/04/2024
Example: Line Balancing
In the production line’s current form (assuming each task is performed on
one WS by one worker),
• What is the bottleneck activity?
• What is the production capacity per hour of the system?
• Redesign the product flow diagram to balance the line and increase the
capacityper hour of the system.
Activity Average timein second
1 20
2 30
3 10
4 10
5 15
6 17
1 2 3 4 5 6
39
Example: Line Balancing
1 2 3 4 5 6
(CT, Cap) = (20, 180) (30, 120) (10, 360) (10, 360) (15, 240) (17, 212)
The bottleneck is: WS2
The current capacity is: 3600/30=120
Total task time = stacked time = 20+30+10+10+15+17 = 102 secs.
We aim for 102/6 = 17 secs.
How can we achieve the cycle time approximately at 17 seconds?
40
20
15/04/2024
Example: Line Balancing
1 2 3 4 5 6
(CT, Cap) = (20, 180) (30, 120) (10, 360) (10, 360) (15, 240) (17, 212)
One person one hour
3
(20, 180)
1 2 5 6
4
(20, 180)
(CT, Cap): (20, 180) (30, 120) (15, 240) (17, 212)
41
Example: Line Balancing
3
(20, 180)
1 2 5 6
4
(CT, Cap): (20, 180) (30, 120) (15, 240) (17, 212)
(20, 180)
Two persons
one hour
2 3
(30, 120) (20, 180)
1 5 6
2 4
(30, 120) (20, 180)
(20, 180) (15, 240) (20, 180) (15, 240) (17, 212)
42
21
15/04/2024
Learning Outcomes
• Longest-Task-time Heuristic
• Kilbridgeand WesterMethod
• Ranked Positional Weights Method
• Compare LTT, STT, KWM, and RPW
• Incremental Utilisation Heuristic
• Other Ways To Improve Line Balance
• Product leveling
• Mixed-model production leveling
• Weighted Average Time method/rule
• Remarks on production levelling
43
Product Leveling
• If we produce more than one product in production
line, what does production Levelling/smoothing
mean?
• Large batches of the same product may reduce set-
up times and changeovers, but usually result in:
– long lead times;
– swelling inventories;
– greater opportunities for defects;
– excessive idle time and/or overtime.
• Product levelling
44
22
15/04/2024
Mixed-Model Production
• Mixed-model production
– Averaging both the volumeand the production sequence
of different model types on a production line.
• Example: Toyota Manufacturing
Toyota makes 3 car models -a convertible, hardtop, and a SUV.
Assume that customers are buying 9 convertibles, 9 hardtops,
and 9 SUVs each day. What is the most-efficient way to make
those cars?
45
Mixed Model Production Leveling
One solutionwould be for Toyota to make all 9 convertibles
in the morning, all 9 hardtops in the afternoon, and all 9 SUVs
in the evening. That would allow people to concentrate on
one kind of work at a time.
.
Uneveness of equipment and people that make different cars over
morning, afternoon, and evening.
46
23
15/04/2024
Mixed Model Production Leveling
In the staging lot, vehicles would pile up between the plant and
the dealers.
Customersdon't buy 9 convertibles in the morning, 9 hardtops
in the afternoon, and 9 SUVs in the evening. They buy different
kinds of cars through the day and week.
Parts Factory
Car Factory Dealer
47
Mixed Model Production Leveling
Toyota solved the problem by production levelling.
• If customers are buying 9 convertibles, 9 hardtops, and 9 SUVs each day,
• Toyota assembles 3 of each in the morning, 3 of each in the afternoon,
and 3 of each in the evening 3A,3B,3C; 3A,3B,3C; 3A,3B,3C;
• It also distributes the production of convertibles, hard tops, and SUVs as
evenly as possible through each shift: convertible, hard top, SUV,
convertible, hard top, SUV, and so on. ABC,ABC,ABC; ABC,ABC,ABC;
ABC,ABC,ABC;
48
24
15/04/2024
Mixed Model Production Leveling
Leveling production also helps to avoid the problem of excess inventory
of finished vehicles. The vehicle plants make the different types of cars
at about the same pace that customers buy those cars. They can adjust
the pace of production as buying patterns change.
As the result, dealersonly need to maintain a minimal inventory of cars
to show and sell.
Parts Factory Car Factory Dealer
49
Mixed Model Production: Example
The monthly customer demands are 4800, 2400, 1200 units for three
products A, B, C, respectively (over 20 days). We want to meet customer
demands and arrange production to be as mixed as possible.
Solution steps:
1. Determine daily production requirements
2. Determine repeating sequence through;
i. Determining the largestINTEGERdivides into the daily volume
ii. Calculate the minimum ratio, e.g. 240, 120, 60 => 60 is the integer,
and so ratios are 4, 2, 1 (60 repeats daily)
iii. Define a product ordering within the repeat sequence
Monthly Days Daily
Prod. A 4800 20 240
Prod. B 2400 20 120
Prod. C 1200 20 60
All 8400 20 420
50
25
15/04/2024
Mixed Model Production: Example
Determine sequence: 240/120/60 → 4/2/1 → AAAABBC → AABABAC
Product ratios: q =240/420=0.571; q =120/420=0.286; q =60/420=0.143;
A B C
Assume 100% efficiency and the available production time per day is 7
hours. Each product requires four tasks with task times in the table.
TaktTime = ? (minutes) for each product and for overall.
For A, 420/240 =1.75; For B, 420/120=3.5;
For C, 420/60=7.0; For all, 420/420 = 1.
Prod. A Prod. B Prod. C
Task 1 0.7 0.6 0.5
Task 2 0.6 0.5 0.6
Task 3 0.5 0.4 0.4
Task 4 0.6 0.6 0.7
Total 2.4 2.1 2.2
Daily Production 240 120 60 total: 420
51
Mixed Model Production: Example
• How to balance: By assigning tasks to a sequence of workstations (WS)
such that:
– The cycle time (CT) for each product at each WS satisfies the Takt
Time (TT) or Required CT
– Tasks are assigned as efficient as possible
• Weighted Average Time (WAT) method/rule:
– WAT is the time required on average at a WS to perform the task
– RULE: WAT at each WS should not exceed the TT
WAT = ∑∑ q x T ≤ TT; i.e. Sum of the product_ratios * task_time < Takt
j i j ij
Where q = proportion of product j in the sequence (q =D/∑ D)
j j j j
T = time for performing task ion product j
ij
For all the tasks that are assigned to each WS.
52
26
15/04/2024
Mixed Model Production: Example
Prod. A B C
Task 1 0.7 0.6 0.5
Task 2 0.6 0.5 0.6
Task 3 0.5 0.4 0.4
Task 4 0.6 0.6 0.7
Total 2.4 2.1 2.2
Daily Production 240 120 60 total: 420
Product ratios: q =0.571; q =0.286; q =0.143;
A B C
For all products, TT= 420/420 = 1min.
WS Task Weighted Average Time (WAT) in minutes
1 1 0.571(0.7) + 0.286(0.6) + 0.143(0.5) = 0.643
2 2,3 0.571(0.6+0.5) + 0.286(0.5+0.4) + 0.143(0.6+0.4) = 1.029
3 4 0.571(0.6) + 0.286(0.6) + 0.143(0.7) = 0.614
53
Mixed Model Production: Example
Workstation WS1 WS2 WS3
Task 1 2+3 4
A 0.7 1.1 0.6
A 0.7 1.1 0.6
B 0.6 0.9 0.6
A 0.7 1.1 0.6
B 0.6 0.9 0.6
A 0.7 1.1 0.6
C 0.5 1.0 0.7
Sum 4.5 7.2 4.3
Average 0.643 1.029 0.614
TT=1min; a sequence = seven products; Every 7 minutes, we produce 4A, 2B and 1C.
This will repeat 60 times per day (assume 7 hours per day).
54
27
15/04/2024
Mixed Model Production: Example
WS1 WS2 WS3
Task 1 2 3 4
A 0.7 0.6 0.5 0.6
A 0.7 0.6 0.5 0.6
B 0.6 0.5 0.4 0.6
A 0.7 0.6 0.5 0.6
B 0.6 0.5 0.4 0.6
A 0.7 0.6 0.5 0.6
C 0.5 0.6 0.4 0.7
Sum 4.5 6.8 4.7
Average 0.64 0.97 0.67
55
Some Remarks
• Production Levelling (Heijunka) points to a knowledge-driven
processfor ensuring stability, flow and pull
• Most lean operations strike a balance between product
levelling and volume levelling
• It also needs to strike a balance between having inventory
and meeting demand
• PL requires a lot of data, and can be tough to deal with
• PL is not necessarily much useful for businesses with level and
dependable demand.
56
28
15/04/2024
Production Levelling in Logistics
• Milk Run Distribution versus Production Levelling
• Milk Run is a delivery method for mixed loads from different
suppliers planned as cyclically repeated tours.
• Instead of each of several suppliers (e.g. 5) sending a vehicle
every week to meet the weekly needs of a customer, one
vehicle visits each supplier on daily basis and picks up
deliveries for that customer.
• Get its name from dairy industry where one tanker collects
milk every day from several farmers to a milk processing firm
Meyer, A (2015). Milk Run Design: Definitions, Concepts and Solution Approaches, Scientific Publishing.
57
Individual vs Milk Run Delivery
www.nipponexpress.com/about/csr/environment/cooperation.html
58
29
15/04/2024
Comments on Milk Runs
• In Japan, supplier MR are widely used for connecting regular
suppliers with the corresponding plants;
• In Germany (and Europe), MR for supplier-plant-relations are
rarely taken.
• MR allow to ship parts more frequently and in smaller lots;
• Regular schedules reduce the variability of the processes and
increase the planning ability in the inbound network
• Suppliers: reduce their safety stocks;
• Logistics providers: improve reliability and reduce empty
return trips;
• Society: reduce emissions and congestion
Meyer, A (2015). Milk Run Design: Definitions, Concepts and Solution Approaches, Scientific Publishing.
59
Discussion: Benefits and Implications
Milk run in logistics is a process for inbound deliveries to warehouses. These deliveries can involve internal or
external supply chains.
Example 1. There are many different growers in one region supplying wheat to a mill that makes flour. Each
farm could send a delivery truck with a load of grain to the mill each week (or day). That’s a quite standard SC
operation. In milk run logistics, a truck would go from the mill to the farms. At each stop on the route, the truck
picks up wheat from a different grower. If each farmer only has a partial load to deliver, this is a more efficient
way to collect the grain.
Example 2. In a heavy equipment manufacturing facility, several sections of the facility fabricate different parts
of the machinery. Another team does the assembly. Someone from each fabrication department could deliver
finished pieces to the assembly section. In milk run, a driver is sent to the different sections to load and
transport all the parts to the assembly area. In this situation, one driver takes the place of many, and each
department’s workforce can focus on fabrication.
A milk run can be a more efficient way to handle logistics. But they do take planning. If the route involves
products from different companies, you need an agreement about cost-sharing and other aspects of the
cooperative delivery arrangement. Once the group settles these issues, this delivery method can save time and
money for everyone by pooling operation costs and resources.
Receiving operations also benefit from milk runs. In the example of the mill, instead of processing a dozen
deliveries from different farms, the mill can get a full wheat delivery from just one truck. In the example of
internal facility, the assembly operations only have one delivery to keep track of rather than running down
parts deliveries from multiple fabrication departments.
efficiency; planning; coordination vertically and horizontally; inventory; time; cost;
utilisation; tracking delivery; competitors.
60
30
15/04/2024
Further Reading: Applications of Line
Balancing
• Koltai, T., Tatay, V. & Kallo, N. (2014) Application of the results of simple
assembly line balancing models in practice: the case of a bicycle
manufacturer, International Journal of Computer Integrated
Manufacturing, 27, 887-898
• Sime, H., Jana, P. and Panghai, D. (2019). Feasibility of Using Simulation
Technique for Line Balancing In Apparel Industry, Procedia
Manufacturing, 30, 300-307.
• Bongomin, O., Mwasiagi, J.I., Nganyi, E.O. and Nibikora, I. (2020).
Improvement of garment assembly line efficiency using line balancing
technique, Engineering Report, 1-18.
61
Summary
• Longest-Task-time Heuristic
• Kilbridgeand WesterMethod
• Ranked Positional Weights Method
• Compare LTT, STT, KWM, and RPW
• Incremental Utilisation Heuristic
• Other Ways To Improve Line Balance
• Product leveling
• Mixed-model production leveling
• Weighted Average Time method/rule
• Remarks on production levelling
62
31