24th International Forum on COCOMO – November 2009
Innovative design Manufacturing
Research Centre (IdMRC)
“World-leading research in engineering
design, manufacture and verification.”
Cost Modelling
Linda Newnes, Head of Costing Research
University of Bath
Estelle Huang (Bath), Glenn Parry (BBS, UWE),
Ricardo Valerdi (MIT)
Presentation Outline
• Present the IdMRC what we do.
• Overview of the work undertaken at Bath in
cost modelling.
• Costing for availability
–
–
–
–
Product Costing
Availability Contracting
Service Costing
Costing for Availability
24th International Forum on COCOMO – November 2009
Innovative design and Manufacturing Research Centre
(IdMRC)
• One of 16 specialised research centres.
• Focus of the activity at Bath is the integration of
Design and Manufacturing.
• IdMRC leads a 3 year Grand Challenge in through
life information and knowledge (KIM project –
Knowledge Information Management).
The vision is to be an internationally leading
research Centre in the synthesis of design,
manufacture and product verification
24th International Forum on COCOMO – November 2009
IdMRC – Four themes to realise our vision
• Constraint Based Design and Optimisation.
(CBDO)
• Advanced Machining Processes and
Systems. (AMPS)
• Metrology and Assembly Systems and
Technology. (MAST)
• Design Information and Knowledge (DIAK)
24th International Forum on COCOMO – November 2009
Through Life Costing
• Our focus is on concept design through to disposal.
• Emphasis on knowledge information and
management for cost modelling.
• Importance of ‘servitisation’ cost modelling (PSS).
• Current collaborators include for example; BAE
Systems, Airbus UK, GE Aviation, Ministry of Defence
(Defence Equipment and Support) and Spirax Sarco.
The overall aim is to provide methods and tools
for managing TLC from concept design to
in-service/disposal.
24th International Forum on COCOMO – November 2009
Why is cost modelling important
• NAO (2008) highlights key areas
– MOD projects late delivery and over budget
– 20 largest projects on average 96 months late and £205M
over budget
• Deloittes (2008)
– Cost overruns in defence and aerospace in next 10 years
26% increase
– Equates to 46%.
• Bernard Gray - MOD Defence Acquisition (2009)
– Average programme overruns by 80% (5 years)
– Approx £300M overrun (frictional cost £900M-2.2bn/annum)
– Underestimating, capability, cost models for strategic
decisions.
– DE&S spend £12bn p.a. on equipment availability
24th International Forum on COCOMO – November 2009
Through Life Costing
Feedback to inform future design
24th International Forum on COCOMO – November 2009
Some cost modelling techniques
• Synthetic Cost Modelling Here the ‘experts’ do their
best ‘guestimate’ on what the Cost Estimating
Relationships (CER) are.
• Generative Cost Modelling (as design progresses
detail improves and cost models e.g. material,
processes etc – lots of detail required).
• Parametric Cost Modelling (using past knowledge
to predict cost e.g. weight of material used in
aerospace and injection moulding). In otherwards
you find a CER.
24th International Forum on COCOMO – November 2009
Uncertainty in cost modelling
• Different options:
– Sell the product and spares
– Lease the product (computers)
– Availability contracting
• Future contracting of some aerospace/defence
products – availability contracts.
How do you cost for through life availability?
24th International Forum on COCOMO – November 2009
Why is the utilization phase important?
Up to 75%
24th International Forum on COCOMO – November 2009
Costing for Availability – current modelling
• Cost modellers have focused on costing products.
• Commercial systems are in general still product
based.
• Little evidence of in-service/utilisation modelling.
• However product and services debated since 1776
by Smith*
• Availability contracts already in place, however
benefitted with transition between products and
service.
*Smith, Adam.
(1776). The Wealth of Nations, Books I-III, Chichester: Wiley.
24th International Forum on COCOMO – November 2009
Key Changes in the Definition of Goods (G) & Services
(S)
Smith (1776)
clarified labour
in terms of
productive (e.g.
Goods) and nonproductive
(e.g.Services)
Smith (1776)
identified unique
G&S
characteristics
Say (1803)
first introduced
the concept of
Materialability
Araujo & Spring
(2006)
The current trend is
to integrate
manufacturing into
service to provide
solutions throughout
the lifecycle of the
product
Senior (1863)
classified G
as an object
and S as a
performance/
act
Axelsson & Wynstra
(2002)
Rather than
distinguishing services
and goods explicitly, it
would be more helpful to
organising & managing
firms with complex
product-service offerings
Hicks (1942)
Identified another
key G&S
characteristics
Shostack (1977)
among the first to
argue that
intangibility can no
longer be the
distinction to
separate S from G
Hill (1999)
among the first
to recognised
the ambiguity to
separate S from
G based on
Perishability
Delaunay&
Gadrey (1987,
cited in
Araujo&Spring,
2006)
formed produceruser interaction as
the basis to
distinguish
between G&S
24th International Forum on COCOMO – November 2009
Aim of availability research
To provide an approach or solution to the challenge
of modelling and analysing the provision of throughlife costing for a service
• To analyse the differences and similarities between
product cost estimating techniques and service cost
estimating techniques
• To design and evaluate an appropriate cost model for
product service systems by identifying in-service
activities and applying estimating rules
• To provide a framework for the provision of service
through-life-costing
24th International Forum on COCOMO – November 2009
Challenge
How do you cost for through life availability or
capability?
• How can industry estimate the TLC for their
products? In particular,
– How can you predict the in-service (utilization/operations
support) costs for the products at the concept design stage
to enable informed decision making?
– Model the cost of decisions e.g. last time buy – how many
parts are stored in warehouses from last time buys
• Uncertainty modelling to aid decision making
24th International Forum on COCOMO – November 2009
Through Life Cost
• Decision making for the acquisition,
development and ongoing support of complex
engineering systems with extended life
– Most meaningful at early stage but highly
uncertain
– Quantitative and objective estimating of TLC
taking into account of uncertainty in estimate
– Use uncertainty modelling for decision making
through the supply chain
24th International Forum on COCOMO – November 2009
The Framework
24th International Forum on COCOMO – November 2009
Uncertainty and Decision Making
• Aleatory uncertainty
– Irreducible randomness associated with the physical system
or the environment
– e.g. repair time, failure rates
– Decision making under risk
• Epistemic uncertainty
– Reducible uncertainty due to a lack of knowledge of
quantities or processes of the system or the environment
– e.g. future decisions
– Decision making under uncertainty
• Important to make a distinction
– Different theories and methods to model these uncertainties
– Can be updated as knowledge is accumulated
24th International Forum on COCOMO – November 2009
Current Approach
• Probabilistic cost risk analysis
– Three-point estimate is common, most likely, min and max to
describe triangular distributions
– Subjective probability is often used to represent uncertainty
• Commercial software incorporate probabilistic
modelling technique
– Monte Carlo with various types of distributions (e.g. uniform,
triangular, normal)
Cost Model
24th International Forum on COCOMO – November 2009
Cost Models Uncertainty
More important to characterise epistemic uncertainty
Estimation Methods
Sources of Uncertainty
Uncertainty to
Characterise
Intuitive/expert opinion

Judgement

Epistemic
Analogical

Selection of benchmark
model (qualitative
characteristics)

Epistemic
Parametric





Cost drivers/parameters
CER choice
Goodness-of-fit
Data uncertainty
Extrapolation

Epistemic and
aleatory
Analytical/engineering



Scope
Level of details
Available data

Epistemic and
aleatory
Extrapolation from
actual costs


Changes in conditions
Limited data

Epistemic and
aleatory
More important to characterise aleatory uncertainty
24th International Forum on COCOMO – November 2009
Cost Data Uncertainty
Data
Uncertainty
Source
Type
Example
Variability
Inherent
randomness
Aleatory
Repair time, Mean Time Between
Failure, labour cost.
Statistical errors
Lack of data
Epistemic
Reliability.
Vagueness
Linguistic
uncertainty
Epistemic
The component may need to be
replaced every 2 to 3 months.
Ambiguity
Multiple sources
of data
Epistemic
Expert 1 and expert 2 provides
different values to end-of-life
costs.
Subjective
judgement
Optimism bias
Epistemic
Over confidence in schedule
allocation.
Imprecision
Future decision
or choice
Epistemic
Supplier A or B.
24th International Forum on COCOMO – November 2009
Scenario Uncertainty
• A scenario is a conceptual model that is developed
(with assumptions) to approximate the actual TLC of
the artifact
– new technologies or legislation, supply chain disruptions,
design changes
• Typical scenarios are the
most likely, worst and best
cases, and the outcomes
of these are then used as
3-point estimates
• No distinction between risk
and uncertainty
Probability density (1/£)
£
Optimistic
Most likely
24th International Forum on COCOMO – November 2009
Pessimistic
Imprecise probability
• When both variability
and imprecision is
present, e.g. uncertain
about the distribution
parameters e.g.
reliability
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
Normal
Mean
St Dev
Upper
bound
12
2
Lower
bound
20
2
Probability
Bounds
0
0
0
5
10
15
PDF
20
25
30
35
0
5
10
15
20
CDF
24th International Forum on COCOMO – November 2009
25
30
35
Probability vs Imprecise Probability
There is 90% probability that
in service cost is within £23M.
F(x)
_
F(x)
There is 90% probability that
cost is within £22M-£24M.
 Use this information to set
contingency budget, best case
£22M worst case £24M.
F(x)
24th International Forum on COCOMO – November 2009
Conclusions
• Uncertainty important in the context of
decision making in TLC
• Separating epistemic and aleatory uncertainty
in costing
– Can update the model when imprecision is
reduced/eliminated, e.g. deciding on alternative
– Allows reuse of objective data e.g. repair time
independent of the probability of repair events
• More transparent decision making under
uncertainty and risk
24th International Forum on COCOMO – November 2009
References
Goh, Y.M, Newnes L.B., Mileham A.R.,
McMahon, C.A and Paredis, C. A framework
for considering uncertainty in quantitative life
cycle cost estimation. ASME 2009, San
Diego, 30th August – 2nd September 2009.
24th International Forum on COCOMO – November 2009
Current Requirements
• We know the literature and approaches of
commercial systems.
• We have some industrial views of how they
model in-service/utilisation costs.
• Need further industrial feedback.
• Complete questionnaires to ascertain the
type of modelling you undertake.
24th International Forum on COCOMO – November 2009
24th International Forum on COCOMO – November 2009
Innovative design Manufacturing
Research Centre (IdMRC)
“World-leading research in engineering
design, manufacture and verification.”
Any Questions