Each LGO student completes a six month internship project at one of our 27 partner companies. The projects range from energy and sustainability, lean operations, supply chain, and many more. All contain significant research which then becomes the basis for the thesis.
1. MIT Leaders for Global
Operations Internship Projects
Fall 2015
2. LGO Internships
One of the hallmarks of the LGO program is the six month internship project,
completed at one of our 27 partner companies. Internship projects vary broadly in
industry and function, but all contain a significant research component which then
becomes the foundation of the student’s thesis.
Included in this slide deck are eight internship projects from the Class of 2016.
3. Table of Contents
National Grid
Risk Analysis of Unmanned Aircraft Systems in National Airspace for Utility
Applications (Jackee Mohl)
Page 5
Amazon.com
Inventory Management for Throughput Optimization (Jake Stowe)
Page 11
Calibra
Finesse Commercialization, Supply Chain Development, and Automation
Implementation (Samer Haidar)
Page 25
Massachusetts General Hospital
Routine Post-Procedure Recovery (RPPR) Patients (Kfir Yeshayahu)
Page 29
4. Table of Contents Continued
Pacific Gas & Electric
Cathodic Protection Resurvey Process – Natural Gas Distribution System (Lillian
Meyer)
Page 43
Amgen
Streamlining and Standardizing Transcriptomic Analysis in Process Development
(Kerry Weinberg)
Page 56
Nike
Leveraging Consumer Sales Data (Blair S. Holbrook)
Page 70
Raytheon
Additive Manufacturing of Metals (Andrew Byron)
Page 78
5. LGO internship project by Jackee Mohl, LGO ’16
(MBA and SM in aeronautics and astronautics)
National Grid:
Risk Analysis of Unmanned Aircraft Systems in
National Airspace for Utility Applications
Company Sponsors: Mike McCallan, Kara Morris
Faculty Advisors: Georgia Perakis, Woody Hoburg
National Grid (Mohl) – 1 of 6
6. Objective - Operational
Operational Goals
• FAA approval for UAS operations
• Instructions for pilot program
• Safety, training and operations plans
• Inspection access and safety
• Helicopter operations cost
and schedule
ProblemProblem SolutionSolution
• Implementation of UAS in
utility operations
National Grid (Mohl) – 2 of 6
7. Approach - Operational
Pilot Program DevelopmentPilot Program DevelopmentFAA
Approval
FAA
Approval
Source: R4 Robotics
National Grid (Mohl) – 3 of 6
8. Objective – Regulatory Research
Regulatory Research Goals
• Recommendations for loosening of restrictions
• Additional data for utility pushback to FAA
• Recommendations for additional safety technology or
restrictions required
• FAA regulations too
restrictive
• Limits potential use cases
ProblemProblem SolutionSolution
• Probabilistic risk analysis
model to determine
equivalent level of safety
National Grid (Mohl) – 4 of 6
10. Status and Next Steps
Next Steps
• Further research data and gain access required for models
• Run probability models and define recommendations
• Further develop pilot program
FAA Section 333 Exemption Ground Collision Model Midair Collision Model
National Grid (Mohl) – 6 of 6
11. LGO internship project by Jake Stowe, LGO ’16
(MBA and SM in engineering systems)
Amazon.com:
Inventory Management for Throughput
Optimization
Company Sponsors: Brian Donato and Joanna Hicks
MIT Faculty Advisors: Dr. Bruce Cameron and Dr. Roy Welsch
Amazon (Stowe) – 1 of 14
12. Context
• In 2012 Amazon purchases Kiva Systems Inc.
• Amazon has built several fulfillment centers reliant on Kiva
technology
• Kiva dramatically changes production model of warehouse
operations
Legacy FC (LFC)
Labor Constrained
Robotic FC (RFC)
Station Constrained
Amazon (Stowe) – 2 of 14
14. Fundamental Questions
• Cost structure of RFCs is substantially less than LFCs
• Natural question: How do we increase throughput of a
Robotics FC?
• A few things to focus on:
– Capacity – More stations? More pods? More floor space?
>> Expensive and complex
– Utilization – Increase stow rates, increase pick rates
>> Hard
– Inventory – Are we stocking the right stuff in the right
places, in the right way?
Amazon (Stowe) – 4 of 14
15. How Inventory Flows in a RFC
Robotic Prime Field
(“Fast”)
Robotic Prime Field
(“Fast”)
Reserve Racks
(“Slow-ish”)
Reserve Racks
(“Slow-ish”)
ReceiveReceive PackPack
ShipShip
Amazon (Stowe) – 5 of 14
16. Further Fundamental Questions
• What kind of inventory are we storing in our
robotic field?
• How fast does it move?
• Is it getting older, or younger on average?
• How should we deal with slow moving
inventory?
• Why does it matter?
Amazon (Stowe) – 6 of 14
17. What kind of inventory are we storing?
• Deadwood – Older than 90 days – not projected to be
sold for 6 months.
Amazon (Stowe) – 7 of 14
19. What to do about it?
Robotic Prime Field
(“Fast”)
Robotic Prime Field
(“Fast”)
Reserve Racks
(“Slow-ish”)
Reserve Racks
(“Slow-ish”)
ReceiveReceive PackPack
ShipShip
Amazon (Stowe) – 9 of 14
20. What to do about it? – Right Now
Robotic Prime Field
(“Fast”)
Robotic Prime Field
(“Fast”)
Reserve Racks
(“Slow-ish”)
Reserve Racks
(“Slow-ish”)
ReceiveReceive PackPack
ShipShipStore slow moving ASINs in
modified reserve racks
Replen when demand is generated
Amazon (Stowe) – 10 of 14
21. What to do about it? – Ideal State
Robotic Prime Field
(“Fast”)
Robotic Prime Field
(“Fast”)
Reserve Racks
(“Slow-ish”)
Reserve Racks
(“Slow-ish”)
ReceiveReceive PackPack
ShipShip
Amazon (Stowe) – 11 of 14
22. Why Does it Matter?
• If you don’t increase capacity or utilization, why
should this matter?
• Pile-on and Pick Density
Legacy Robotic
Amazon (Stowe) – 12 of 14
23. Next Steps
• Create a realistic model of how consolidation
activities will affect utilization
• Use of optimization techniques to determine an
optimal number of station hours to devote to
consolidation
• Pilot and scale up storage in the reserve racks
(beginning this week)
• Larger Impact for Industry
– Greater understanding of the dynamics of
inventory in automated warehousing operations
– Exploring the applications of legacy vs. automated
operations
Amazon (Stowe) – 13 of 14
24. Key Challenges and Interests
• Challenges
– Complexity of the systems and stakeholder interests
– Siloed knowledge about the technology and human
processes
• Three major knowledge areas are necessary:
– Human Processes: Leadership of the FC and line
Workers
– Kiva Technology: Kiva engineers and systems
engineers tasked with optimizing system
– LGO Knowledge – Time series analysis and
mathematical optimization
Amazon (Stowe) – 14 of 14
25. LGO internship project by Samer Haidar ’16
(MBA and SM in mechanical engineering)
Calibra:
Finesse Commercialization, Supply Chain
Development, and Automation Implementation
Company Sponsors: Eijiro Kawada, Hector Rodriguez, Jim Conroy
MIT Faculty Advisors: Dimitris Bertsimas, Brian Anthony
Calibra/Johnson & Johnson (Haidar) – 1 of 4
26. Objective
What I am hoping to learn:
•How to manage a 20-member cross-functional project team within a very large
healthcare company
•How to design and optimize a customer-centric end-to-end supply chain for a new
product launch
Background
•Calibra was a venture-backed startup that Johnson & Johnson acquired in 2012.
•The Calibra Finesse is a wearable insulin delivery patch that will be launched
in 2016. It delivers 2 units of bolus insulin per click to help diabetic patients
control blood glucose at mealtime.
Objective and Benefits
•The internship will develop a 7-year strategic plan for a lean, end-to-end supply chain for the Calibra
Finesse.
Calibra/Johnson & Johnson (Haidar) – 2 of 4
27. Approach
Approach
Using the Process Excellence tools at J&J and the Lean/Six-Sigma training we received at LGO,
we are following the DMADV process design framework:
Resources that I will need
•Guidance from faculty advisors and project leader and champion
•Functional leaders within project team to define current state processes and metrics
•Gartner supply chain benchmarking insights
-Project Charter
-Identify & segment
customers
-Voice of Customer
-Define metrics
-Current State Analysis
-Benchmarking
-High level
design/modelling
-Develop detailed
design elements
-Develop validation
and control plans
-Verify against business case
-Validate against VOC
Calibra/Johnson & Johnson (Haidar) – 3 of 4
28. Status and Next Steps
Status Update
•Collected and analyzed Voice of Customer (VOC) data from interviews and surveys with persons with
diabetes, healthcare professionals, wholesale distributors and retailers.
•Translated VOC needs into supply chain’s functional requirements.
•Developed modelling framework for manufacturing facility location-allocation problem.
Findings
•Customer journey mapping reveals critical steps in insulin device training, ordering, use, and support.
•Affinity mapping of VOC needs statements uncover strategic supply chain design requirements focused
on visibility, collaborative planning, and order management.
Next steps
•Develop current state value stream map highlighting different functions’ processes.
•Benchmark best-in-class companies based on critical functional requirements.
•Develop future state design concepts and evaluate against functional requirements.
•Expand manufacturing facility location-allocation model to incorporate supplier selection and distribution
scenarios.
Calibra/Johnson & Johnson (Haidar) – 4 of 4
29. LGO internship project by Kfir Yeshayahu ’16
(MBA and SM in electrical engineering and computer science)
Massachusetts General Hospital:
Routine Post-Procedure Recovery (RPPR)
Patients
MGH Sponsors: Dr. Peter Dunn, Bethany Daily, Cecilia Zenteno
MIT Faculty Advisors: Retsef Levi, Patrick Jaillet, David Scheinker
MGH (Yeshayahu) – 1 of 14
30. Massachusetts General Hospital
• MGH annually handles ~1.5 million
outpatient visits and admits
~48,000 inpatients
• More than 37,000 surgeries
annually, in 70 Operating Rooms
spread across five buildings on
three floors Dr. John Collins Warren performs the first surgery
without pain as William Morton administers ether
““Since 1811, Massachusetts General Hospital (MGH) has been committed toSince 1811, Massachusetts General Hospital (MGH) has been committed to
delivering standard-setting medical care.”delivering standard-setting medical care.” (MGH website)(MGH website)
MGH (Yeshayahu) – 2 of 14
31. Routine Post-Procedure
Recovery (RPPR) Patients
Surgical patients categories:
• RPPR is an internal category in MGH and is not used by insurance companies.
• In practice, a hospital bed is being reserved for all RPPR patients.
MGH (Yeshayahu) – 3 of 14
32. Project Objectives & RPPR
Challenges
Project Goal: Determine optimal patient flow strategies for RPPR patients
by analyzing the trade-offs in resource utilization of alternative pathways.
RPPR Challenges:
MGH (Yeshayahu) – 4 of 14
34. Status
• Studied RPPR booking considerations:
– bed assignments, procedure types, insurance reimbursement, etc.
• Generated a Recovery Flow Charts for certain surgical
procedures
• Created a Length of Stay comparison tool (web application)
• Analyzed patterns over time
– Example: Implications of approach change after the Lunder building
opened
MGH (Yeshayahu) – 6 of 14
35. Status
• Studied RPPR booking considerations:
– bed assignments, procedure types, insurance reimbursement, etc.
• Generated a Recovery Flow Charts for certain surgical
procedures
• Created a Length of Stay comparison tool (web application)
• Analyzed patterns over time
– Example: Implications of approach change after the Lunder building
opened
MGH (Yeshayahu) – 7 of 14
37. Status
• Studied RPPR booking considerations:
– bed assignments, procedure types, insurance reimbursement, etc.
• Generated a Recovery Flow Charts for certain surgical
procedures
• Created a Length of Stay comparison tool (web application)
• Analyzed patterns over time
– Example: Implications of approach change after the Lunder building
opened
MGH (Yeshayahu) – 9 of 14
38. Length of Stay Comparison App
MGH (Yeshayahu) – 10 of 14
39. Status
• Studied RPPR booking considerations:
– bed assignments, procedure types, insurance reimbursement, etc.
• Generated a Recovery Flow Charts for certain surgical
procedures
• Created a Length of Stay comparison tool (web application)
• Analyzed patterns over time
– Example: Implications of approach change after the Lunder building
opened
MGH (Yeshayahu) – 11 of 14
40. Study Different Approaches: Keep in the
PACU vs. Send to Floors
Length of stay definition:
from entering the PACU to
discharging from the hospital
Patient population: RPPR who
were discharged home
directly from the PACU
Time frame: CY2010-CY2013
Data sources: Operating
Rooms & Admitting
department (CBED)
Legend:
Keep in the PACU (Before 09/2011)
Send to Inpatient Floors
Legend:
Keep in the PACU (Before 09/2011)
Send to Inpatient Floors
MGH (Yeshayahu) – 12 of 14
41. Next Steps
• Map & acquire additional necessary data to analyze
wasted bed-time
• Model alternative mechanisms of handling RPPR
patients in terms of recovery location and surgery
scheduling
• Create a predictive model for patients’ length of stay
in the hospital
• Generate operational recommendations
MGH (Yeshayahu) – 13 of 14
42. Top 20 RPPR Procedures in 2014
MGH (Yeshayahu) – 14 of 14
43. LGO internship project by Lillian Meyer ’16
(MBA and SM in civil and environmental engineering)
Pacific Gas & Electric:
Cathodic Protection Resurvey Process – Natural
Gas Distribution System
Company Sponsors: Preston Ford, Sumeet Singh, Mallik Angalakudati
MIT faculty Advisors: Herbert Einstein (CEE), Georgia Perakis (Sloan)
PG&E (Meyer) – 1 of 13
44. Gas Operations within PG&E
Source: "Natural Gas System Overview." Natural Gas System Overview. Pacific Gas & Electric Company.
PG&E (Meyer) – 2 of 13
45. PG&E approaching industry trend
Source: USA. U.S. Department of Transportation. PHMSA. Distribution, Transmission &
Gathering, LNG, and Liquid Annual Data.
PG&E (Meyer) – 3 of 13
46. Direct cost of corrosion is $276
billion per year in US
Infrastructure,
$22.6
Transportation,
$29.7
Productionand
manufacturing,
$17.6
Government,
$20.1
Electrical utilities,
$6.9
Gas distribution,
$5.0
Drinking water and
sewer systems,
$36.0
Utilities, $47.9
Cost of Corrosion in billions
Source: Corrosion Costs and Preventive Strategies in the United States. McLean, VA: Federal Highway
Administration, 2002. NACE International.
Note: Cost of corrosion includes monitoring, replacing, and maintaining
PG&E (Meyer) – 4 of 13
47. Prioritizing a system-wide survey
based on risk
• How do we properly allocate resources? Where can
we start that eliminates the most risk?
• Goal: a systematic way of designating where leaks
due to corrosion are most likely to occur throughout
PG&E’s distribution pipeline system
PG&E (Meyer) – 5 of 13
49. Significant amount of work still to
be completed
• Currently
– Gathering data about the current pipeline system
– Researching relevant corrosion models
• Next steps
– Perform regression analyses and compare predictive factors
to historical leak data
– Develop risk-ranking model and identify high-risk areas
– Perform system-wide survey based on risk analysis; were
high-risk areas addressed first?
PG&E (Meyer) – 7 of 13
50. 3 key challenges to consider
• Sheer size of PG&E’s network
• Communication between different databases and
data storage methods over time
• Sustainability of maintenance practices
PG&E (Meyer) – 8 of 13
53. PG&E has an increasing number
of corrosion leaks repaired
Source: USA. U.S. Department of Transportation. PHMSA. Distribution, Transmission &
Gathering, LNG, and Liquid Annual Data.
PG&E (Meyer) – 11 of 13
54. Identifying probability and
consequence of failure
Source: Ogosi, Eugene, and Stephen McKenny. "Indexing Model for Pipeline Risk
Assessment and Corrosion Management." NACE Corrosion 2014 (2014):
OnePetro.
PG&E (Meyer) – 12 of 13
56. LGO internship project by Kerry Weinberg ’16
(MBA and SM in bioengineering)
Amgen:
Streamlining and Standardizing Transcriptomic
Analysis in Process Development
Company Sponsors: Brian Follstad (Amgen supervisor), Sam Guhan (Amgen champion)
MIT Faculty Advisors: Doug Lauffenburger, Roy Welsch
Amgen (Weinberg) – 1 of 14
57. Agenda
• Amgen biotech process development
• Current complex data analysis requires streamlining
workflow and developing internal tools
• Current Status: Beta version of tool developed
• Next Steps: Application of beta tool to historical datasets
• Key challenges and opportunities
• Q&A
Amgen (Weinberg) – 2 of 14
58. Amgen Process Development
R&DR&D PDPD
CHO cell
ProductivityProductivity
QualityQuality
Ribosome
making
therapeutic
protein
Ribosome
making
therapeutic
protein
Correct
sugars on
therapeutic
protein
Correct
sugars on
therapeutic
protein
(Image sources in notes)
MfgMfg
Amgen (Weinberg) – 3 of 14
59. Amgen Process Development
• Process Development group improves and
characterizes bioreactor productivity and final product
quality by analyzing the cellular impact of process
conditions
Productivity1
F1(a,b,c)F1(a,b,c)
CHO Bioreactor
A
B
C Quality1
Productivity2
F2(a,b,c)F2(a,b,c)
Quality2
Cellline1Cellline2
A
B
C
Amgen (Weinberg) – 4 of 14
60. Transcriptomic Analysis in Amgen
Process Development
Transcriptomic Data Collection
Transcriptomic Data Analysis
F1(a,b,c)F1(a,b,c)
CHO Bioreactor
A
B
C
Productivity1
Quality1
Pathway
Analysis
Pathway
Analysis
ClusteringClustering
Principle
Component
Analysis
Principle
Component
Analysis
Amgen (Weinberg) – 5 of 14
61. Transcriptomic Analysis in Amgen
Process Development
+
Understanding CHO
Biological System
Transcriptomic Data Statistical and
Pathway Analysis ProductivityProductivity
QualityQuality
Current transcriptomic data analysis workflow
Ad-
hoc
Ad-
hoc
Time
consumin
g
Time
consumin
g
Steep
learning
curve
Steep
learning
curve
Ad-
hoc
Ad-
hoc
Amgen (Weinberg) – 6 of 14
63. Approach
Assess current
state workflow
Assess current
state workflow
Develop
automated
tool
Develop
automated
tool
Refine tool
and develop
standard work
Refine tool
and develop
standard work
PilotPilot
Knowledge
Transfer
Knowledge
Transfer
CHO transcriptomic
Database
(Amgen)
Information
Systems
Literature
review &
advisors
Feedback from
core users
Feedback from
core users
Agile
sprints
Agile
sprints
Process dev
R&D
Publically available
data
(chogenome.org)
Amgen (Weinberg) – 8 of 14
64. Status and Next Steps
Assess current
state workflow
Assess current
state workflow
Develop
automated
tool
Develop
automated
tool
Refine tool
and develop
standard work
Refine tool
and develop
standard work
PilotPilot
Knowledge
Transfer
Knowledge
Transfer
Publically available
data
(chogenome.org)
Information
Systems
Literature
review &
advisors
Feedback from
core users
Feedback from
core users
Agile
sprints
Process dev
R&D
Agile
sprints
CHO transcriptomic
Database
(Amgen)
Amgen (Weinberg) – 9 of 14
67. Is this sustainable?
• Open source software code
• Tool designed for future extensions
• Highly documented source code
• Process development investing in Python knowledge
Amgen (Weinberg) – 12 of 14
68. What is the business impact?
• Simpler data analysis workflow drives future use of
omic analysis
• Cost of omic analyses ~$1k vs. Cost of shutdown
bioreactor run ~$1M
• Understanding link between process inputs and
process outputs = key for comparability
Amgen (Weinberg) – 13 of 14
70. LGO internship project by Blair S. Holbrook ’16
(MBA ad SM in engineering systems)
Nike:
Leveraging Consumer Sales Data
Company Sponsors: Jon Frommelt and Mike Overson
MIT Faculty Advisors: Tauhid Zaman and David Simchi-Levi
Nike (Holbrook) – 1 of 8
71. Company Context
CORE
ESSENTIALS
LONG
LIFECYCLE
1 WEEK ORDERS
SHORT LEAD
TIME
SEASONAL
INNOVATIVE
6 MONTH ORDERS
CUSTOMIZED
PERSONALIZED
2 WEEK ORDERS
QUICK
RESPONSIVE
3 MONTH ORDERS
SEASONAL QUICK TURN
ALWAYS
AVAILABLE CUSTOM
Nike (Holbrook) – 2 of 8
72. Problem and Key Question
How do we reduce bullwhipping?
CUSTOMERNIKE ALWAYS AVAILABLE CONSUMER
Nike (Holbrook) – 3 of 8
75. Findings To Date
POS based forecasts can
be much more accurate.
…yet current forecasts
can be highly inaccurate .
POS data can be
consistent…
Some products are better
suited for forecasting.
Nike (Holbrook) – 6 of 8
76. Status and Next Steps
1. Select style-colors
2. Develop statistical forecasts
3. Socialize and secure pilot commitment
4. Develop reorder policy and initiate
pilot
5. Measure outcomes
Nike (Holbrook) – 7 of 8
77. Key Challenges
1. Securing commitment from partner – Remember
Barilla?
2. Executing
3. Scalability – Not overpromising
Nike (Holbrook) – 8 of 8
78. LGO internship project by Andrew Byron ’16
(MBA and SM in aeronautics and astronautics)
Raytheon:
Additive Manufacturing of Metals
Company Sponsors: Manuel Gamez, Teresa Clement
MIT Faculty Advisors: Steven Eppinger, Brian Wardle
Raytheon (Byron) – 1 of 5
79. Why develop additive
manufacturing?
• Raytheon produces precision
weapons using advanced
manufacturing
• Additive manufacturing (AM)
can help develop and
produce complex products
faster
• Industry has begun to define
practices and characteristics,
but most is primary research
or proprietary
• Aerospace applications need
reliability and repeatability
Image source: Raytheon Product Information, Standard Missile-3
Courtesy of TWI Ltd
Image source: Wikimedia
Raytheon (Byron) – 2 of 5
80. How to drive change
• We are developing a
business process programs
and factories can use to
ensure qualified, predictable
results for AM parts
• Part of that effort involves
understanding the effect of
process inputs – materials,
controls, procedures
– Initiated through a designed
experiment on primary metals
AM control parameters
Process
Capability
Material
Specification
Equipment
Qualification
NDT Qualification
Integration,
Verification &
Validation (IV&V)
Statistical Sample
Testing
Inspection Process Control
Design Feasibility
Part Development
Plan
Proof Parts
Preliminary
Statistical
Characterization
Main Factors Units
Laser Power W
Scan Speed mm/s
Scan Spacing µm
Beam Diameter µm
Feedstock Factors
Flowability Go/NoGo
Dose Factor %
Reuse/Recycle Cycles #
Particle Size (sieving) µm
Coater Blade Height µm
Post-Process Factors
Hot Isostatic Pressing Y/N
Responses Units
Dimensional Accuracy %
Surface Finish RMS
Density (vs. wrought) %
Layer Cohesion 1-5
Ultimate tensile strength ksi
Yield tensile strength ksi
Ductility (Elongation) ε
Melt Pool Diameter µm
Raytheon (Byron) – 3 of 5
81. • First level of qualification process nearly developed
– Builds required, testing and estimated schedule
Availability 100
Units 1
AM Specialist
AM Build
Cycle Time N(18, 0.6)
Mold
Breakout/
Powder
Cleanup
Cycle Time U(2.5, 0.3)
Welding
Cycle Time U(1.1, 0.8)
Is Part
Welded?
Surface
Finishing
Cycle Time U(2, 0.5)
HIP
Cycle Time U(3, 0.2)
(Exit)
Assembly
Yes
No
Loading
Cycle Time 0.6
Is Feedstock
Available?
Feedstock Lot
Prep
Cycle Time T(2, 3, 5)
Baseplate
Removal
Cycle Time U(3, 0.75)
Inspection
Cycle Time 0.6
Availability 100
Units 1
Machinist
Status and next steps
ProModel Process Simulation diagram of model used to represent AM parts build process and duration
Raytheon (Byron) – 4 of 5
82. • Initial screening experiment complete
Next Steps:
• Final product will be a recipe for any material or
process: a set of steps to qualify AM for a new part
• Experimental results will be analyzed and used to
create a more specific factored test on targeted
parameters
• End goal is to develop a process “recipe” that can be
used for any material or AM technology
Raytheon (Byron) – 5 of 5
Editor's Notes
Background about MGH and the collaboration team
What is RPPR?
Why is RPPR an issue?
How are we going to attack the issues?
What was done so far?
What was done so far?
Example of analysis we did (recovery flows).
Mention that some flows are bad, and this is a way to catch them.
What was done so far?
Example of some of the work itself – not an actual result. Fits here?
What was done so far?
Example of an analysis we did (effects of the PACU approach)
What are we going to do next?
Introduce project, myself, and advisors at PG&E and MIT
10 minutes for presentation
Introduce PG&E and its distribution pipelines
In a straight line, gas network is SFO to BOS 15 times
Steel distribution is SFO to BOS 6.75 times
Walk through graph.
Discuss increasing number of corrosion leaks and self-reports of unprotected pipe to the CPUC. These two factors indicate that a survey to re-evaluate the system is necessary.
Mention controls from corrosion (first coatings required; then in 1970s, CP required)
Direct costs includes monitoring, replacing, and maintaining
Transmission pipelines included in infrastructure, not utilities ($7 billion)
Transition from costs to project
State problem goal within this CP survey.
Looking at the factors that indicate the probability of failure (ie the probability that corrosion will occur) and the consequence of failure (population density, public image, cost). Both influence risk.
Corrosive environment – soil resistivity, soil pH, moisture content
Corrosion control conditions – impressed current or galvanic? Age, condition of coating
History of pipe – material, historical installation procedures, age, pipe always protected from corrosion
Previously have met with stakeholders. Currently in the complicated data-gathering process / exploring how other groups in the industry are evaluating risk with respect to corrosion.
Discuss next steps.
Sheer size of PG&E’s network PG&E’s service area is larger than 33 states; specifically, do we divide over 18 divisions or 4,000+ CP areas?
Quality and accuracy of data in addition, data located in different systems undergoing transitions
Quality control how do we make sure that CP continues to be documented accurately and promote communication between departments so that no additional unprotected pipe is accidentally created?
Conclude by stating ultimate goals of this survey and risk model are 1) allocate resources to drive down risk in high risk areas and, as a result, 2) future results of dramatic decrease in corrosion leaks.
Note that graph beyond heavy black line is idealized projection of future results
Open up the floor for Q&A (approx 2 min)
Discuss increasing number of corrosion leaks and self-reports of unprotected pipe to the CPUC. These two factors indicate that a survey to ensure system is in full compliance is necessary.
Looking at the factors that indicate the probability of failure (ie the probability that corrosion will occur) and the consequence of failure (population density, public image, cost). Both influence risk.
Conclude by stating ultimate goals of this survey and risk model are 1) allocate resources to drive down risk in high risk areas and, as a result, 2) future results of dramatic decrease in corrosion leaks.
Open up the floor for Q&A (approx 2 min)
Note that patterned part of graph is idealized projection of the results that we would like to see.
Motivation behind this project, what’s the objective of this project and my approach. Then describe where I’m at in the project and some next steps. Close with some key challenges and interests thus far and then some q&a
Be sure to explain mammalian cell culture process dev or biotech process dev
Drug substance
Need to also mention the “knobs” you can adjust on a genetic level to the cells. i.e. overexpression of Mgat to impact high mannose. Essentially engineering cell lines to maximize productivity (not just using process conditions to impact it).
Also mention that optimizing this process is key for companies that are investing in biosimilars where their cost is primarily associated with cost of production rather than cost of R&D
Notes: spell out: cell line & production vessel, bioreactors,
Cell line + process conditions, process conditions impact the environment, cell line is the function. F(a,b,c), f_amg123, f_amg345.
Images sources described in earlier slide notes.
Verbally explain what the impact of this project is to Amgen (streamlining workflow to reduce time spent in analysis), identification of bio signatures import for productivity to improve bioreactor productivity. For academia: demonstration of analysis standardization for transcriptomic data, identification of biological signatures indicative of productivity (relation between process conditions and CHO cellular activity).
Documentation of workflow,
Figuring out how people analyzed something can be difficult
Ultimate outcome: shorten time to market?
Link to : shorten time to market, less lot to lot variability, more consistent product quality to focus effort on key specifications (i.e. for patients)
-patient sensitivity to different levels of product quality (differences in glycosylation)
Show images for each point-> i.e. code, show automation, “leverage automated workflow”: impact of variety of steps & preprocessing sensitivity (look at old work)
Merge 3 & 4 together: and focus on application of tool & workflow
Agile development: (sprints) include
Show image of tool. Need to include significant findings so far, How to move forward to achieve project goal,
CHO transcriptomic databases: CHO genome.org
Knowledge Transfer: include R&D as well
Main story: standard mamm cell culture process: goals are productivity & quality, need ways to control that. Transcriptomics provide understaniding, problem: this analysis is difficult/ad hoc/complex to figure out, so this solution is to provide a standard/traceable clear way of doing the analysis by automating it (creating code) vs. enterprise software tool that has many options vs. etc.
Goals, omics fits in with goals, current state of omics analysis, solution ( can verbally close with this)
Data structure: could be applied to other industries
Verbally explain what the impact of this project is to Amgen (streamlining workflow to reduce time spent in analysis), identification of bio signatures import for productivity to improve bioreactor productivity. For academia: demonstration of analysis standardization for transcriptomic data, identification of biological signatures indicative of productivity (relation between process conditions and CHO cellular activity).
Documentation of workflow,
Figuring out how people analyzed something can be difficult
Ultimate outcome: shorten time to market?
Link to : shorten time to market, less lot to lot variability, more consistent product quality to focus effort on key specifications (i.e. for patients)
-patient sensitivity to different levels of product quality (differences in glycosylation)
Show images for each point-> i.e. code, show automation, “leverage automated workflow”: impact of variety of steps & preprocessing sensitivity (look at old work)
Merge 3 & 4 together: and focus on application of tool & workflow
Verbally explain what the impact of this project is to Amgen (streamlining workflow to reduce time spent in analysis), identification of bio signatures import for productivity to improve bioreactor productivity. For academia: demonstration of analysis standardization for transcriptomic data, identification of biological signatures indicative of productivity (relation between process conditions and CHO cellular activity).
Documentation of workflow,
Figuring out how people analyzed something can be difficult
Ultimate outcome: shorten time to market?
Link to : shorten time to market, less lot to lot variability, more consistent product quality to focus effort on key specifications (i.e. for patients)
-patient sensitivity to different levels of product quality (differences in glycosylation)
Show images for each point-> i.e. code, show automation, “leverage automated workflow”: impact of variety of steps & preprocessing sensitivity (look at old work)
Merge 3 & 4 together: and focus on application of tool & workflow
Write out top few questions and have backup slides
Write out top few questions and have backup slides
Write out top few questions and have backup slides