Truck set-up scheduling for the Elgin, IL facility

Vijay Viswanathan, Graduate StudentDr. Jaejin Jang, Associate Professor

Department of Industrial Engineering

Overview of Presentation

• Introduction• Detailed Problem Description• Objective of Project• Assumptions• The Scheduling Algorithm• Simulation Input• Summary of Results• Conclusion

Introduction

• Wisconsin Lift Truck corporation sells, rents, services trucking equipment like…– Aerial lift and

construction equipment– Cranes– Lift Trucks – Industrial Cleaning

Equipment– Loaders – Pallet Trucks – Personnel Carriers

Introduction

• They have seven locations in Wisconsin and two locations in Illinois

• Our contact for this project is – Kristine Rauch

• Industrial Engineer, WLT, Brookfield, WI

Receive Truck

Stock Sold to Customer

Inventory WIP

Schedule for Set-Up

Stage

Delivery(Corp run or

Branch truck)

Set-Up(Bay Area)

Problem Description

• Bay has maximum capacity of 25– 5 stations with capacity of 5 trucks each

• 4 are considered WIP • 1 truck is being worked on

• 5 techs available for scheduling– one tech assigned to each of the 5 bays

Problem Description Cont’d

• 4 set-up types• 19-Used Truck Setup

– 3 types of New Truck Setups• 20-N.T SET-UP• 36-NEW RTL SET-UP• 76-I/C NT SETUP

Tech 1

Tech 2

Tech 3

Tech 4

Tech 5

Schematic of the Truck Set-up Bays

Scheduling Objective

• Minimize the time it takes for stock and new customer truck set ups to get from WIP to the delivery area

• Minimize tardiness to ensure timely delivery of trucks to customers as much as possible

Assumptions• There will be no preemption during

setup, once a job is started, it is completed – If we cannot finish a truck without

stopping in the middle of processing, we do not start the truck

• Once a tech has completed a setup they will start the next set up right after

• Both stock and customer trucks are taken together as number of stock trucks is negligible

• 6 hour work day

Receive Truck

Schedule for Set-Up

Truck set for

delivery

Set-Up(Bay Area)

Scope for scheduling

• 2 areas where scheduling can be done– When picking trucks from the infinite

arrival area into one of the 5 WIPs– When selecting one truck for set-up in

each bay station

Raw Data Analysis• Data From: Jan 2005 – May 2006 i.e.; 16

months– Job Type 76 – 515 jobs– Job Type 36 – 43 jobs– Job Type 19 – 97 jobs– Job Type 20 – 449 jobs– Total Significant Jobs: 1104

• % of Job Type 76 – 46.65%• % of Job Type 43 – 3.89%• % of Job Type 19 – 8.79%• % of Job Type 20 – 40.67%• Total No. of Trucks - 1974• No. of Stock Trucks [IMH RENTAL] - 89

[4.5%]• No. of Sold Trucks – 1885 (95.5%)

Simulation Input Data

–Setup times for each job type• Total labor $$ billed on each truck available• Tech billed at $97 per hour

–Distributions for inter-arrival times, by job type

• Used ARENA Input Analyzer module to fit each of these into standard probability distributions

More Simulation Inputs..

• Due dates were assigned as an attribute to every job arriving– Due date distribution also determined

from invoice data

• Results were used to design an ARENA simulation so that different scheduling algorithms could be implemented and their performance evaluated.

Scheduling LOGIC

• Experimented with 4 different hybrid scheduling algorithms.

• Scheduling Algorithms used are a simple combination of standard algorithms with modifications made to suit this problem.– LST and SPT

• Process trucks in both WIP and bay queues in order of lowest value of the sum of Slack time and Processing time.

• In doing so, we anticipate minimizing flowtime as also reducing slack.

– LST OR SPT• A modification of the above algorithm where incoming trucks

are processed as per the least slack time, if (slack >0), else SPT is used to select trucks from the queues.

Scheduling Logic

– EDD With Priority• Modification of the EDD algorithm to incorporate a

priority mechanism for new truck types over old truck types

• Selection rule is – least value of (due date/priority), where priority is “50” for a new truck type and “10” for a used truck type (type 19)

– Shortest Processing Time • Plain ol’ SPT in least order of processing time

Simulation run Parameters

• Replication Length – 180 Days• Hours/day – 6 hrs• Warm up period – 10 Days• Number of Replications -4

Simulation Results [ALL UNITS DAYS]

JOB TYPE 20

RULE Combination Avg. Flowtime Max. Flowtime Avg. Lateness Max. Lateness

LST_SPT 5.33 19.48 517.44 1832

LST_OR_SPT 5.53 31.75 638.97 1681.74

EDD_PRIO 7.34 41.28 393.22 1705.47

SPT 7.09 59.86 422.55 1507.55

JOB TYPE 19 [USED TRUCK]

RULE Combination Avg. Flowtime Max. Flowtime Avg. Lateness Max. Lateness

LST_SPT 2.65 16.14 406.02 1104

LST_OR_SPT 5.84 31.12 466.45 1373.9

EDD_PRIO 7.38 29.55 501.92 1550.34

SPT 4.87 56.84 618.11 1527.85

Conclusion• The combination of LST and SPT works best to

minimize overall flow time (but not tardiness).• If either LST/SPT is used depending on

whether the job is slack or not, this leads to a significant reduction in max. lateness, compared to using LST + SPT in all cases.

• SPT scheduling yields the least max. lateness.• A priority based scheme predictably decreases

the avg. lateness for new truck set-ups. • Future Work Idea

– An Intelligent scheduler that can decide on the right scheduling algorithm based on a reasonable guess of whether incoming jobs will be late or not.

Thank you!

Questions/Comments?