Slide 1 Randolph E. Kirchain 1 , Elsa A. Olivetti 1 , Melissa L. Zgola 1 , Christopher Weber 2 , Sarah Boyd 3 , Ramzy Kahhat 4 , Eric Williams 4 1 Massachusetts Institute of Technology, 2 Carnegie Mellon University, 3 University of California at Berkeley and 4 Arizona State University Environmental Assessment of IT Products Development of the PAIA Tool
28
Embed
Environmental Assessment of IT Productsthor.inemi.org/webdownload/newsroom/Presentations/EnviroLeaders... · Slide Slide 4 4 Meeting the Need of ICT for Quantification Identifying
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Slide 1
Randolph E. Kirchain1, Elsa A. Olivetti1, Melissa L. Zgola1,
Christopher Weber2, Sarah Boyd3, Ramzy Kahhat4, Eric Williams4
1Massachusetts Institute of Technology, 2Carnegie Mellon University, 3University of California at Berkeley and 4Arizona State University
Environmental Assessment of IT ProductsDevelopment of the PAIA Tool
Slide 4Slide 4
Meeting the Need of ICT for QuantificationIdentifying the Challenges
• Data availability
– Process Data
• Specialized materials and processes
• Depth / dynamics of the supply chain
– Product Data
• Number / diversity of products
• Complexity and dynamics of product
• "Distance" between designer and impact
• Cost / Speed• Data collection and analysis
• Action-ability• Relating design and operations decisions to
product impact
• Credibility• Significant uncertainty exists in estimates; must
be acknowledged and accommodated
Clear need: Efficient &
effective
approach
to LCA
Slide 5Slide 5
Meeting the Need of ICT for QuantificationTranslating the Challenges to Project Objectives
• Goals
– Develop near term, quantitative approach for labeling
• Resolve product types (13” vs 15” not 4310 vs 6410)
• Provide insight into major sustainability levers
– Create breadboard tool
• Objectives for approach / tool
– Efficient
• Minimum data collection & Minimum user input
– Effective
• Resolve product type
• Provide actionable insight
– Credible
• Uncertainty is explicitly estimated
• Transparent & Flexible
Slide 6Slide 6
Project Strategy:Realizing Efficient / Effective IT LCA
Two major strategies to meet goals
1. Probabilistic Triage and Progressive Refinement
2. Product Attribute to Impact Algorithm (PAIA)
– An approach that maps product attributes to their environmental impact
Initial focus
– Product: Laptop
– Impacts: Energy & carbon
Slide 7Slide 7
Realizing Quantitative Streamlined LCA:Tradeoff between Comprehensiveness and Specificity
Comprehensiveness
Specificity
Idealized
Goal
Results precise
Resource intensive
Omissions indefensible
Results accurate
Targeted resources
Significant uncertainty
Screening
Slide 8Slide 8
Examples of Sources of Uncertainty
• Data availability concerning
– Bill of materials (laptop vs. particular model)
– Suppliers practices and location
• Representativeness of secondary data
– Variation in supplier technology
– Geographic variation
• Grid mix, efficiency, transportation
– Temporal variation
• Process and process evolution
Uncertainty may be resolvable at an acceptable cost
Slide 9Slide 9
Supplier Derived Variability:Significant Variation Exists in Real-world Suppliers
International Aluminum Institute 2003
Slide 11Slide 11
Project Strategy:Realizing Efficient / Effective IT LCA
Two major strategies to meet goals
1. Probabilistic Triage and Progressive Refinement
2. Product Attribute to Impact Algorithm (PAIA)
– An approach that maps product attributes to their environmental impact
Initial focus
– Product: Laptop
– Impacts: Energy & carbon
PAIA Streamlines Process
Slide 12Slide 12
Product Attribute to Impact Algorithm (PAIA): The Basic PAIA Concept
• Minimum user input, attributes which are
– Important
• Significant effect on results
• Viewed as critical by stakeholder
– Knowable (Measurable at “low” cost)
Inputs Results
Product Type Attributes
Laptop
15 “ Screen
250 GB Hard drive
6 Layer
PWB
…
Product Type Impacts
MJ Energy
Kg CO2
Gal H2O
…
Product Attribute to
Impact Algorithm
Slide 13Slide 13
Attribute-
to-Activity
Model
Impact
Assessment
Model
BOA
LCI
Inventory
Database
Existing LC tools
Realizing the Product Attribute to Impact Algorithm
Incorporating Engineering Models with Existing Tools
Inputs Results
Product Type Attributes
Laptop
15 “ Screen
250 GB Hard drive
6 Layer
PWB
…
Product Type Impacts
MJ Energy
Kg CO2
Gal H2O
…
Activity Amount
Aluminum 20 g
Electricity 140 KWh
Lithography 0.5 g
Inj. molding 40 g
Transport 4 tkm
Activity Amount
Anhydrite, in ground 0.1 kg
Carbon dioxide, in air 1.2 kg
Oil crude, in ground 3.6 g
Land Transformation 40 km2
Zinc, in ground 0.2 kg
Total g Al = a*LCDsize + b*HDD capacity
Total g PC = c*Chassis + d*PWB area
g Lithography = e* layers PWB +
d*Ictype
Etc…
Numbers are for illustration only
Slide 15Slide 15
Overall Triage Approach to Creating PAIA
1. Leverage existing data to create best available estimate
– Gather existing BOA and LCI data
– Assemble uncertainty information
• LCI database mining and data reduction
• Manufacturing and grid market data
• Government or third party usage studies
• Extreme conditions(e.g., rail vs. truck)
2. Develop & execute LCI simulation (Monte Carlo) model
3. Triage (screen) for high impact activities
4. Develop PAIA modules to relate attributes to activities– Assemble training LCAs
– Create correlative models
Slide 16Slide 16
Developing the Laptop PAIA
Manufacturing
Logistics
PackagingUse
EoL
Slide 17Slide 17
Comprehensive Probabilistic Triage:
Analysis breakdown by LC phase
95% of statistical trials indicate that
90% of the impact attributed to
Matls & Mfg and Use phase * transport phase
-50
50
150
250
350
450
550
650
GW
P (
kg C
O2
-e
q)/
pro
du
ct c
lass
Overall
Coefficient
of Variation ~35%
Slide 18Slide 18
Comprehensive Probabilistic Triage:
Analysis of Components (Matls & Mfg.)
95% of trials indicate that 75%
of the impact attributed to
LCD, Mainboard, and Chassis
0
50
100
150
200
250
300
GW
P (
kg C
O2
-e
q)/
pro
du
ct c
lass
Slide 21Slide 21
Targeted analysis around LCD:
Each resolved driver lowers COV
0
0.1
0.2
0.3
0.4
0.5
Unresolved Screen size
PFC abatement
Location Energy intensity
IC/PWB area
Co
eff
icie
nt
of
Vari
ati
on
Slide 22Slide 22
Current set of overall model inputs:
Main drivers of impact
• Manufacturing Context– Location and efficiency
• LCD– Size
– PFC abatement
– Bulb technology
– IC/PWBs
• Mainboard– IC impact node, chip area,
integration, PFC abatement
– PWB area
• Chassis– Materials (Mg vs. Plastic)
– Inner cages
• Hard drive– Capacity
– Technology (SSD-future tech)
• Battery – Number of cells
• Transportation –assembly to customer
– Mode, distance
• Packaging*– Mass and recycled content
• Use phase– Duty cycle, power, grid,
lifetime
*More a hot button issue than a hot spot
Slide 23Slide 23
Current model screenshot
Front page example, includes inputs/levers
Example only, not actual results
Slide 24Slide 24
-50
50
150
250
350
450
550
650
Targeted assessment:
Results in lower overall variation
Targetedassessment
Overall COV ~ 10%
Comprehensive assessment
Overall COV ~ 35%
-50
50
150
250
350
450
550
650
GW
P (
kg C
O2
-e
q)/
pro
du
ct c
lass
Slide 25Slide 25
Project Accomplishments
• Project accomplishments
– Triage and targeted refinement identifies
• important inputs for user
• Important focus for data refinement
– Mapping attributes to impacts is possible / promising
• Limited levers account for majority of variation
• Continuing work– Testing resolution of current Laptop model
– Extending to Desktops and
– Deeper look into LCD Modules
– Primary data collection
– Harmonize with existing efforts
Slide 26Slide 26
Lessons learned
• Uncertainty is significant
• Triage (screening) is still possible
– Limited levers account for much variation
– Data collection can be focused
– Limited analytical data requirement should be possible