Project #3: Production Cost Savings at Washington
Description: Project 3: Production Cost Savings at Washington Post SYST 798OR 680 Final Presentation 6 May 2011 Team Power Wash Post Thomas Kuklinski Timothy Smith Ling Wu Vladimir Zivkovic 1 Overview Clients and Introduction Background, Objective,
Related Topics
Download Presentation
"Project #3: Production Cost Savings at Washington" is the property of its rightful owner. Permission is granted to download and print the materials on this website for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Presentation Transcript
slide1. Project #3: Production Cost Savings at Washington Post SYST 798/OR 680
Final Presentation
6 May 2011
Team Power Wash Post
Thomas Kuklinski
Timothy Smith
Ling Wu
Vladimir Zivkovic 1<br>
slide2. Overview Clients and Introduction
Background, Objective, and Scope
Technical Approach
Model Architecture
Results and Analysis
Recommendations
Future Work
Acknowledgements
Questions 2<br>
slide3. Clients and Introduction Kent Renk, Materials Handling Foreman, renkk@washpost.com
Kim Hammett, Assistant Superintendent for Materials Handling, hammettk@washpost.com 3<br>
slide4. Background In the Fall of 2010, GMU students did initial static analysis of the Washington Post Shipping and Receiving Department in Springfield, VA
They conducted a process evaluation, an analysis of labor, and an analysis of routes which resulted in three groups of recommendations:
Labor reductions (implemented)
Route simplifications
Improved data tracking 4<br>
slide5. Objective and Scope Objective
Provide a baseline simulation model that helps the Washington Post maximize the efficiency of the materials handling processes
Scope
Build upon the static analysis of the previous group by modeling the helper component of the Shipping and Receiving Department
Deliver a flexible simulation model that can be used by the Post to make workforce planning decisions 5<br>
slide6. Technical Approach Conduct a System Process Analysis
Build upon last group’s work done on process evaluation
Develop a complete and detailed qualitative understanding of how the Shipping and Receiving Department operates
Collect and Analyze Data
Collect data from both the materials tracking system (i.e. MTC) and subject matter experts
Obtain descriptive statistics on each component of the system
Select and Construct Model
Evaluate model alternatives based on quantitative/qualitative data and client needs
Construct and test model (e.g. turn system processes into Arena modules)
Perform Output Analysis
Evaluate output in terms of helper utilization, moves, and total production
Iterate 6<br>
slide7. Model Architecture Process Analysis Overview
General Arena Module
Helpers zones and arrivals
General Process Modules
Daily Insert/Sunday Packaging raw ads from rack to machine processing
Completed Daily Insert pallets from SLS 1-6 to CSLD/NDSL
Completed Sunday Packaging from Collators 1-4 to rack/NDSL
Completed Sunday Packaging from rack to CSLD
Schedule Module
Output Analysis Module 7<br>
slide8. Process Analysis (actual) Trucks Materials Handling 2a. Daily Insert Ads 1. Daily Insert Ads & Sunday Packaging 2a. Sunday Packaging Run of Press 2c. Jackets 3a. Completed Jackets 3b. Collated Sunday Packaging 8<br>
slide9. Process Analysis (scope) Materials Handling Input: Sunday Packaging Output: Completed Jackets Output: Collated Sunday Packaging Run of Press Output: Head Sheets Input: Daily Insert Ads Output: Collated
Sunday Packaging 9<br>
slide10. General Module C1HS S1HS Dock Rack TRIA(2,5,6) TRIA(6,8,9) EXP(5) Raw Materials Monday, Shift 1 Production Targets C1PZ C1PZ EXP(5) 10<br>
slide11. General Process Modules Utilizes “Transporter” modules and non-stationary Poisson Process schedule
Internal business logic dictated by historical statistics (e.g. CSLD vs. NDSL) and internal business logic (e.g. Z-Loader selection) 11<br>
slide12. Schedule Module Utilizes “Halt” and “Activate” modules 12<br>
slide13. Output Analysis Modules Utilizes “Record” and “ReadWrite” modules
Records current markup
Records utilization time by helper by shift/day
Records helper moves by helper by shift/day
Records total pallets moved through model by week 13<br>
slide14. Model Architecture (Assumptions) Raw materials start at rack (95% go to rack first)
Exponential/Poisson arrival processes
Triangular route time and Z-Loader distributions
Machine and rack processes as terminate and arrive modules
Internal business logic
NDSL vs. CSLD; NDSL vs. Rack
Least busy Z-Loaders machines are used
Unlimited trucks available for delivery of completed materials
1 week reflects all weeks 14<br>
slide15. Model Architecture (Limitations) Lack of route time data
Plenty of arrival data
Changing work center business logic makes it hard to keep up
Lack of fully developed business logic for the internal model process (e.g. 50% NDSL vs. CSLD)
Recipes for advertisements and rack waiting times make it difficult to model a seamless transition of materials through the system 15<br>
slide16. Results and Analysis Input Analysis
Labor Analysis
Route Analysis
Arrival Process Analysis
Output Analysis
Base vs. Suggested Schedule
Helper utilization and moves
Total production 16<br>
slide17. Labor Input Analysis Obtained labor markup data for this period
Broken down by work center, day, shift, and role
Assigned helpers within the work centers to specific routes they were responsible for
This required splitting helpers assigned to specific routes into subsets (e.g. Collator – Deep Reach)
Trimmed off helpers in the markup that would be working on tasks not reflected in the model (i.e. Loaders/Unloaders working at the receiving dock) 17<br>
slide18. Route Input Analysis Mapped all of the routes to specific areas within our system
Routes obtained from MTC Pallet Move Summary Report for February 7-12, 2011
Obtained route time and distance for all relevant routes from client
Client provided estimated max, min, and mean routes times for a triangular distribution
Client recommended this approach because of “bad” MTC data
Used route times, distances, and helper assignments to determine velocities of each helper for model input 18<br>
slide19. Arrival Processes Input Analysis Obtained arrival timestamp summaries for components of the system from MTC
Rack Sunday Packaging and Daily Insert raw ads
SLS 1-6 & Collators 1-4 finished pallets
Empty Run of Press jackets
Rack finished Sunday Packaging
Used Arena’s Input Analyzer to determine distributions on the arrival processes for the components based on timestamps
Distributions broken down by shift/day and machine
Exponential inter-arrival times
Poisson arrival process 19<br>
slide20. Output Analysis Output analysis was conducted as a demonstration of how the model could be used to inform decisions
Primary deliverable was usable decision tool going forward
Technical session and model instructions ensured client understanding of the model functionality
Compared base schedule and suggested schedule
Utilization: Percentage of workers busy at any given time
Helper Moves: Moves per helper-hour
Total Production: Total weekly moves
Sensitivity analysis on arrival processes and helper velocities confirmed results 20<br>
slide21. Base Schedule 21 165 total shifts over a 5-day week
6,500 total helper moves
20% utilization
6 moves per helper-hour<br>
slide22. Suggested Schedule 22 75 total shifts over a 5-day week
90 less than base
6,500 total helper moves
Same as base
40% utilization
Double base
12 moves per helper-hour
Double base<br>
slide23. Future Work Build an Excel-based automation of the input data
Develop animation capabilities to help visualize the processes internal to the model
Enhance the output analysis capabilities to capture more specific details about the statistics of the model
Continue to flesh out big assumptions within the model (e.g. machine processing recipes)
Build a simulation-based optimization engine around the model
Production target goals
Schedule variables 23<br>
slide24. Acknowledgements Washington Post
Mr. Kent Renk
Mr. Bill Thompson
GMU Team I’m PRO WasP
Justine Blaho
Catalina Gomolka
Ryan Graziano
Laura Rodriguez Lopez 24<br>
slide25. Questions? 25 - Remove DVM and schedule chart
- Reduced slides on Input Analyzer
- Remove a portion of "future works"
- On slide 5 ( object and need), remove "life dynamic simulation", instead mention it's a baseline simulation model that allows the customer to play with.
- in Technical Approach section slide 8, rewords the four phrases and make them active ( a verb + a noun)
- on slide 16, modify the sentence "z-loaders go to least busy"
- on slide 17, modify "develop" to "developed"
- mention somewhere in the tech secion on the slides, that the custmer accept the model after we did a client-side tech review session.
- remove evaluation page
- add more details on the output analysis sections
- add content on optimizing the model, make it clear optimizing the model is NOT in the sense of "LP", call it "improved model" or "modified model" instead, and eleborate why it's an improved model ( save on labor cost, etc.)
- Add sensitivity analysis section to talk about assumptions that we are less sure of, run the model with different distributions, verify the assumptions may not be correct. mention that we provided the customer a tool which need continuing effots to improve
- Tim and Tom coordinate with each other so that the produciton process will not be repeated twice<br>
Final Presentation
6 May 2011
Team Power Wash Post
Thomas Kuklinski
Timothy Smith
Ling Wu
Vladimir Zivkovic 1<br>
slide2. Overview Clients and Introduction
Background, Objective, and Scope
Technical Approach
Model Architecture
Results and Analysis
Recommendations
Future Work
Acknowledgements
Questions 2<br>
slide3. Clients and Introduction Kent Renk, Materials Handling Foreman, renkk@washpost.com
Kim Hammett, Assistant Superintendent for Materials Handling, hammettk@washpost.com 3<br>
slide4. Background In the Fall of 2010, GMU students did initial static analysis of the Washington Post Shipping and Receiving Department in Springfield, VA
They conducted a process evaluation, an analysis of labor, and an analysis of routes which resulted in three groups of recommendations:
Labor reductions (implemented)
Route simplifications
Improved data tracking 4<br>
slide5. Objective and Scope Objective
Provide a baseline simulation model that helps the Washington Post maximize the efficiency of the materials handling processes
Scope
Build upon the static analysis of the previous group by modeling the helper component of the Shipping and Receiving Department
Deliver a flexible simulation model that can be used by the Post to make workforce planning decisions 5<br>
slide6. Technical Approach Conduct a System Process Analysis
Build upon last group’s work done on process evaluation
Develop a complete and detailed qualitative understanding of how the Shipping and Receiving Department operates
Collect and Analyze Data
Collect data from both the materials tracking system (i.e. MTC) and subject matter experts
Obtain descriptive statistics on each component of the system
Select and Construct Model
Evaluate model alternatives based on quantitative/qualitative data and client needs
Construct and test model (e.g. turn system processes into Arena modules)
Perform Output Analysis
Evaluate output in terms of helper utilization, moves, and total production
Iterate 6<br>
slide7. Model Architecture Process Analysis Overview
General Arena Module
Helpers zones and arrivals
General Process Modules
Daily Insert/Sunday Packaging raw ads from rack to machine processing
Completed Daily Insert pallets from SLS 1-6 to CSLD/NDSL
Completed Sunday Packaging from Collators 1-4 to rack/NDSL
Completed Sunday Packaging from rack to CSLD
Schedule Module
Output Analysis Module 7<br>
slide8. Process Analysis (actual) Trucks Materials Handling 2a. Daily Insert Ads 1. Daily Insert Ads & Sunday Packaging 2a. Sunday Packaging Run of Press 2c. Jackets 3a. Completed Jackets 3b. Collated Sunday Packaging 8<br>
slide9. Process Analysis (scope) Materials Handling Input: Sunday Packaging Output: Completed Jackets Output: Collated Sunday Packaging Run of Press Output: Head Sheets Input: Daily Insert Ads Output: Collated
Sunday Packaging 9<br>
slide10. General Module C1HS S1HS Dock Rack TRIA(2,5,6) TRIA(6,8,9) EXP(5) Raw Materials Monday, Shift 1 Production Targets C1PZ C1PZ EXP(5) 10<br>
slide11. General Process Modules Utilizes “Transporter” modules and non-stationary Poisson Process schedule
Internal business logic dictated by historical statistics (e.g. CSLD vs. NDSL) and internal business logic (e.g. Z-Loader selection) 11<br>
slide12. Schedule Module Utilizes “Halt” and “Activate” modules 12<br>
slide13. Output Analysis Modules Utilizes “Record” and “ReadWrite” modules
Records current markup
Records utilization time by helper by shift/day
Records helper moves by helper by shift/day
Records total pallets moved through model by week 13<br>
slide14. Model Architecture (Assumptions) Raw materials start at rack (95% go to rack first)
Exponential/Poisson arrival processes
Triangular route time and Z-Loader distributions
Machine and rack processes as terminate and arrive modules
Internal business logic
NDSL vs. CSLD; NDSL vs. Rack
Least busy Z-Loaders machines are used
Unlimited trucks available for delivery of completed materials
1 week reflects all weeks 14<br>
slide15. Model Architecture (Limitations) Lack of route time data
Plenty of arrival data
Changing work center business logic makes it hard to keep up
Lack of fully developed business logic for the internal model process (e.g. 50% NDSL vs. CSLD)
Recipes for advertisements and rack waiting times make it difficult to model a seamless transition of materials through the system 15<br>
slide16. Results and Analysis Input Analysis
Labor Analysis
Route Analysis
Arrival Process Analysis
Output Analysis
Base vs. Suggested Schedule
Helper utilization and moves
Total production 16<br>
slide17. Labor Input Analysis Obtained labor markup data for this period
Broken down by work center, day, shift, and role
Assigned helpers within the work centers to specific routes they were responsible for
This required splitting helpers assigned to specific routes into subsets (e.g. Collator – Deep Reach)
Trimmed off helpers in the markup that would be working on tasks not reflected in the model (i.e. Loaders/Unloaders working at the receiving dock) 17<br>
slide18. Route Input Analysis Mapped all of the routes to specific areas within our system
Routes obtained from MTC Pallet Move Summary Report for February 7-12, 2011
Obtained route time and distance for all relevant routes from client
Client provided estimated max, min, and mean routes times for a triangular distribution
Client recommended this approach because of “bad” MTC data
Used route times, distances, and helper assignments to determine velocities of each helper for model input 18<br>
slide19. Arrival Processes Input Analysis Obtained arrival timestamp summaries for components of the system from MTC
Rack Sunday Packaging and Daily Insert raw ads
SLS 1-6 & Collators 1-4 finished pallets
Empty Run of Press jackets
Rack finished Sunday Packaging
Used Arena’s Input Analyzer to determine distributions on the arrival processes for the components based on timestamps
Distributions broken down by shift/day and machine
Exponential inter-arrival times
Poisson arrival process 19<br>
slide20. Output Analysis Output analysis was conducted as a demonstration of how the model could be used to inform decisions
Primary deliverable was usable decision tool going forward
Technical session and model instructions ensured client understanding of the model functionality
Compared base schedule and suggested schedule
Utilization: Percentage of workers busy at any given time
Helper Moves: Moves per helper-hour
Total Production: Total weekly moves
Sensitivity analysis on arrival processes and helper velocities confirmed results 20<br>
slide21. Base Schedule 21 165 total shifts over a 5-day week
6,500 total helper moves
20% utilization
6 moves per helper-hour<br>
slide22. Suggested Schedule 22 75 total shifts over a 5-day week
90 less than base
6,500 total helper moves
Same as base
40% utilization
Double base
12 moves per helper-hour
Double base<br>
slide23. Future Work Build an Excel-based automation of the input data
Develop animation capabilities to help visualize the processes internal to the model
Enhance the output analysis capabilities to capture more specific details about the statistics of the model
Continue to flesh out big assumptions within the model (e.g. machine processing recipes)
Build a simulation-based optimization engine around the model
Production target goals
Schedule variables 23<br>
slide24. Acknowledgements Washington Post
Mr. Kent Renk
Mr. Bill Thompson
GMU Team I’m PRO WasP
Justine Blaho
Catalina Gomolka
Ryan Graziano
Laura Rodriguez Lopez 24<br>
slide25. Questions? 25 - Remove DVM and schedule chart
- Reduced slides on Input Analyzer
- Remove a portion of "future works"
- On slide 5 ( object and need), remove "life dynamic simulation", instead mention it's a baseline simulation model that allows the customer to play with.
- in Technical Approach section slide 8, rewords the four phrases and make them active ( a verb + a noun)
- on slide 16, modify the sentence "z-loaders go to least busy"
- on slide 17, modify "develop" to "developed"
- mention somewhere in the tech secion on the slides, that the custmer accept the model after we did a client-side tech review session.
- remove evaluation page
- add more details on the output analysis sections
- add content on optimizing the model, make it clear optimizing the model is NOT in the sense of "LP", call it "improved model" or "modified model" instead, and eleborate why it's an improved model ( save on labor cost, etc.)
- Add sensitivity analysis section to talk about assumptions that we are less sure of, run the model with different distributions, verify the assumptions may not be correct. mention that we provided the customer a tool which need continuing effots to improve
- Tim and Tom coordinate with each other so that the produciton process will not be repeated twice<br>