multi-order automation project - university of …the amazon fulfillment engine (afe) is the method...
TRANSCRIPT
The Amazon Fulfillment Engine (AFE) is the method by which the multi-unit
orders are combined within an Amazon fulfillment center (FC).
BFI1 FC in Sumner, WA data will be used as a reference for the project.
Current Flow AFE
1. Individual units that are a part of a multi-unit order are batched together into a
yellow tote.
2. Once this tote is full, it is conveyed to the induct station.
3. Induct operator will scan each item into AFE trays, the tray then transfer to the
AFE sorter.
4. The AFE sorter will sort each tray to the rebin wall.
5. The rebin associate scans the item, which triggers a light on the rebin wall
indicating the chute destination for the item.
6. The rebin associate will walk to the chute location to place the item
7. The rebin associate will return to position for the next item.
8. Once a chute contains all of the units for the multi-unit order, the light on that
chute will flash, and the rebin associate will push the entire order to the pack side
of the chute.
9. Empty trays leaving the rebin associate station travel back to the induction
associate creating the closed loop cycle.
In this project Amazon would like us to develop new
concepts for combining and packaging orders with
multiple items.
Our Opportunities
• Takt time in each station is not balanced
• Rebiner has a lot of non-value added time
Our Objectives
• Improve the output rate in the multi-ordering process
• Find most automated way possible without increasing
the number of employees.
• Higher rate = Use less employees to achieve the same rate
• Current Center: Increase total output per employee by 12.8 – 42.8%
• New Center: Increase total output per employee by 20.1 – 42.8%
• Cost Reduction: 19% in the new fulfillment center
• Minimal equipment cost for current fulfilment center improvement.
• Reduce operator’s fatigue and workload
• Faster training for new employees
Findings:
Siripong Somboon, Yi Yang, Rui Wei, Mark Mao, Tuo Liu
Thank you to those at Amazon who
sponsored our project and cooperate
directly with our team.
Martin Aalund
Michael Hill
Akhil Ranka
Amazon FC Employees at BFI1
Special Thanks to:
Christina Mastrangelo
Who guided throughout the whole project
Multi-Order Automation Project
Recommendations
AFE
Figure 2: Overview of AFE processing area
Figure 1: Close up view of rebin area
1
2
1
2
Induct Station
Rebin Station
1000 unit/hour 700 unit/hour
Recommendation:
Automated staffing system is intended to balance the induct station with
the rebin station.
Expected Benefits:
• Find daily optimal takt time balance between the two stations.
• No new equipment needed, low implementation cost.
• Reduce the workload of the area manager.
Figure 3: Current Conveyer Design Figure 4: New Conveyer Design
Figure 5: Simio Model and OptQuest result
Figure 10: Proposed conveyer design
• 2 cases for current arrangement
• One inductors works with two rebiners
(1:2), 1000 UPH to 1400 UPH
• One inductors work with One rebiners
(1:1), 1000 UPH to 700 UPH
• Both combinations will result imbalanced flow.
• The inductors stand and wait for empty
trays from Rebin section or rebiners spend
time placing excessive trays on the
floor to avoid the trays stuck on the
conveyers.
Recommendation:
• Optimal Ratio 2 inductors to 3 rebiners.
• 2000 UPH to 2100 UPH
Expected Benefits:
• Ideal arrangement: 2:3 ratio
between induction and rebin.
• Increase the rate by ~28%
• Reduce NVA for the operators
• No excessive trays movement
• Stable working station
• Less fatigue for the operators
• Reduce cost up to 19% compare to
current conveyer design
Findings:
• Unnecessary body rotations
• Slow the operators down
• Fatigue easily
B R1L1L2 R2
Figure 6: Shows current position of rebiner.
360⁰ rotation for each item.
B R1L1L2 R2
Figure 7: Shows proposed position of rebiner
resulting in less rotation.
Figure 9: Optimal # of compartments for
different demands
Expected Benefits:
• ~12% increase in rate by changing the position
• Better working condition for the operators (Less
rotation)
Recommendation:
• During peak use 2 inductors and 3 rebiners
• During non-peak use 3 inductors and 3 rebiners
Expected Benefits:
• Can optimize the output rate during peak and non-peak
• Reduce NVA time -> Less training, shorter learning curve
• Use less employees to achieve the same rate (reduce
cost)
• Adaptable to different demand
Figure 8: Sample Calculation for the rate.
Findings:
• Rebin wall is only utilized ~50% during
non-peak time.
• Not enough rebin chute during peak
• Has to use side inefficient rebin wall
Recommendation:
• Increase rebin compartment from 5 -> 7
• Use 3 compartment for non-peak period
• Use 7 compartment for peak period
Expected Benefits:
Peak
Non-Peak
Black Friday; Holiday
Weekday
Weekend
0.69 sec /unit 0.26 sec/unit
Rotation
Reduction
360o to 249o /unit 4.5 m to 4.17m /unit
Total Time Save
0.95 sec/unit
Time
Interval
Avg Rebin
Rate (unit/hr)
Increase
in Rate
Peak Time 662 unit/hr -5.4%
Normal Time 784 unit/hr 12%
Non-peak
Time
949 unit/hr 35.6%
1000 unit/hour
1400 unit/hour
1000 unit/hour
700 unit/hour
2000 unit/hour
2100 unit/hour
+ Walking
Reduction
Recommendation:
Current Fulfillment Center
Rebin Ergonomics and Staffing System
• Easy to implement without interruption to current system
Dynamic Rebin Wall
• Gradually implementing in parallel with current system.
New Fulfillment Center
Rebin Ergonomics and Staffing System
Integrated System Solution
350
500 500
333375.5
400
0
100
200
300
400
500
600
Current Fullfilment Center New Fullfilment Center
Avg
Outp
ut
Rat
e P
er
Pers
on
(Unit/H
our)
Rate Comparison between Current system
and the Proposed System
Current Rate (Non-Peak) New Rate (Non-Peak)
Current Rate (Peak) New Rate (Peak)
• No new equipment needed
Normal