MITIP2006, 11-12 September, Budapest
THE AERO-ENGINE VALUE CHAIN UNDER FUTURE BUSINESS
ENVIRONMENTS: USING AGENT-BASED SIMULATION TO
UNDERSTAND DYNAMIC BEHAVIOUR
David BUXTON, Richard FARR, Bart MACCARTHY,
Division of Operations Management, Nottingham University Business School
Jubilee Campus, NG8 1BB, Nottingham, United Kingdom
Email: d.buxton@nottingham.ac.uk
Abstract:
Agent-based modelling is gaining popularity for understanding the behaviour of complex
systems involving interactions of many players or agents. In this paper an agent-based
simulation modelling technique is applied to understand the long term implications of strategy
decisions for an aerospace value chain. The industry has unique elements including new
business models, high levels of collaboration, long product lifecycles and long periods before
positive paybacks are realised. Forecasting market conditions over this long term lifespan is
inherently problematic and adds further complexity when devising a strategy. The model
described includes all the major players and entities in the value chain and their interactions.
Illustrative results are presented to demonstrate how the approach can be used to evaluate
strategy and policy decisions and their operational implications over the long term.
Keywords:
Value Chain, Simulation, Agent-Based, Business models, Aerospace
1. INTRODUCTION
This paper describes the application of a simulation developed using the principles of agent-based
modelling to future aero-engine business environments. In contrast to alternative modelling
techniques, an agent-based approach follows bottom-up principles. The modelling processes
captures individual processes, decision rules, and exchanges of information and materials. Each
agent runs as an independent simulation, interacting with other agents in the system. This concept
allows for the study of how interactions at the micro-level - the agent level - lead to large scale
system behaviour. Agent-based models are comprised of distinct units modelled at a relatively
detailed level, with particular focus placed on capturing boundary interactions and exchanges.
Collective behaviour and interactions create the dynamics of the system [3].
The aim of this work is to understand the implications of the business strategy and business
models adopted in an aero-engine value chain over a long project lifecycle. The value chain
encompasses original equipment supply, aftermarket services and consumables supply. Each
player in the value chain is represented as an agent, allowing detailed capture of individual
business processes, logic, attitudes to risk and responses to changes in the market place.
2. THE AERO-ENGINE MARKET STRUCTURE
The context for the case study is the European Aero engine industry for commercial jets. The
commercial aerospace industry has some unique characteristics, being affected frequently by
government actions and support. This is perhaps most obviously seen within airlines, where many
current airlines have previously been publicly owned national ‘flag carriers’ and, despite the recent
trend for privatization, government protection is still evident. During the most recent downturn
(2000 to 2003) US airlines received government support of $5 billion and yet still made a collective
loss throughout 2001, 2002 and 2003 [2].
Further government influence can be seen through the use of bankruptcy protection. Although less
common in Europe, US airlines have frequently used the bankruptcy protection of ‘Chapter 11’ to
enable restructuring. Without normal competition dynamics to balance supply and demand,
government protected over-capacity tends to remain, as does the decline in seat yields, which
initially led to the financial problems [2]. Over-capacity and low yields lead to instability, and the
MITIP2006, 11-12 September, Budapest
effects of any disturbances within the global economy are felt by the aerospace industry as the
impact cannot be absorbed.
Economic problems for airlines are compounded by the high investment required in airframe and
engine fleets. The typical 7X7 airframe has a list price in the region of $250 million and the engines
a further $35 million [9]. Although such hardware purchases may have typical lifecycles of at least
25 years, the industries susceptibility to global incidents means that airlines investing in new
aircraft are taking a considerable future demand risk.
For the aero-engine OEM and the value chain, the consequence of customer uncertainty and high
risk has placed downward pressure on the purchase price of new hardware. Discounts on list price
have become the industry norm and may be as high as 70%. OEMs therefore become reliant on
aftermarket revenues (spare parts and maintenance) to ensure profitability. OEMs are therefore
cross subsidizing the sale of original equipment with aftermarket sales, and in so doing are
implicitly absorbing some risk from the airline, creating a situation where both the airline and the
OEM are reliant on customer (seat) demand to be profitable.
Accepting this shared risk is now part of the OEM business model; however, to balance the effect,
OEMs now strive for airlines to enter into long term business deals, guaranteeing long term
relationships and aftermarket revenues. This is achieved through deals such as TotalCare (a RollsRoyce term, http://www.rolls-royce.com/service/civil/totalcare/default.jsp), a bundled product
offering of engines, on-going maintenance services and spares supply. Payment is linked to the
usage of the engine on a per engine flight hour (EFH) and per engine flight cycle (EFC) basis. This
differs to the historic norm for aftermarket sales, when work is undertaken on a ‘time and materials’
basis charged per incident and shop visit. For the airline, the risk is much reduced by TotalCare as
costs are linked to engine usage and the OEM is prepared to share this risk as there is a significant
benefit from creating a long term relationship that will guarantee aftermarket revenues.
Industry dynamics are further complicated by the relationship adopted between OEMs and
upstream value chain partners where risk is also mitigated through Risk and Revenue Sharing
(RRS) agreements. This has the dual benefit of promoting inter-company collaboration by linking
reward to the success of the project. It also shares the risk incurred in the development costs of a
new engine programme (these can be as high as $500 million). The RRS agreements are
structured with the value chain partner ‘buying’ a proportion of the programme and receiving a
percentage share of the revenues generated. As such any changes in revenue flow have
implications for the entire value chain.
These concepts highlight some of the complexity of managing an aerospace business; the
implications of business decisions are difficult to evaluate and, coupled with the long term nature of
paybacks and product lifecycles and the uncertainty of future demand, it is difficult to made truly
informed decisions.
3. TOOLS FOR STRATEGY DECISION SUPPORT
Grant [3] considers business strategy as a theme adopted by a company to give coherence and
direction to the actions and decisions of an organisation’s management. He argues that
implementing a business strategy is not a matter of intuition or good fortune and a number of
concepts and methods can be applied to formulate effective strategies. In practice however, many
of these tools are conceptual, providing a way to visualise a strategy and ‘imagine’ potential
scenarios, but there are few that attempt to play a strategy out over the longer term [7].
Simulation is a proven technique that has been applied frequently within the operations
management discipline through the application of either discrete event simulation or system
dynamics. Discrete event simulation is event and process orientated and therefore lends itself to
capturing high detail level systems, such as manufacturing or service processes [7]. System
dynamics take a ‘systems thinking’ perspective to understand the complex dynamics produced
through interactions of feedback and control mechanisms at a high ‘tree top view’. The concept can
MITIP2006, 11-12 September, Budapest
be used to understand how the high level system behaves and undergoes state changes [7]. It has
been used previously within the aerospace sector to understand the business cycle [6].
System dynamics concepts are frequently applied to strategy problems, well known successful
applications include General Motors and Shell [7]; in both cases system dynamics gave decision
makers the opportunity to ‘play out’ the implications of their decisions and understand that the
optimum short term decision, where analytical techniques can work well, may not be best possible
over the long term. However, system dynamics works at the aggregate level, using rates of change
over time, to show how complex feedback mechanisms can combine to produce unexpected
results. It would be difficult to apply system dynamics to this value chain issue. It might be possible
to model how the market could evolve but it would be very difficult, if not impossible to capture a
detailed understanding of the behaviour of different players in the market place.
Agent-based models have the potential to overcome this problem, each agent may have elements
of discrete event or system dynamics within them, but each is built as a self contained entity
responding to external inputs. The computer based simulation developed for this problem models
value chain performance over an engine development programme. The agents modelled include
the prime partner, customers, the 1st tier Risk and Revenue sharing partners, the 2nd tier materials
suppliers, and the engine (the engine is modelled as an agent, allowing the variability in the
aftermarket to be captured). Each of these agents displays bespoke characteristics and
behaviours.
4. USING AGENT-BASED SIMULATION FOR STRATEGY DECISIONS
The model has been built to demonstrate that it is possible to understand dynamic behaviour of the
market and evaluate how each agent will perform under a future business environment. To achieve
this, a mechanism is needed to allow for the development of realistic business scenarios within
which a strategy can be evaluated. Scenario building allows managers to consider potential
changes in a range of industry drivers and then build a ‘future business environment’ Error!
Reference source not found.. To facilitate this within the aerospace sector the Virtual Interactive
Business Simulator (VIBES) has been previously developed as a visual scenario capture tool [3].
By linking this to an agent-based modelling technique, it will be possible to build and explore future
business environments under a range of parameters, structures and operating policies.
The top level hierarchy for the model is shown in Figure 1.
Figure1. Top level description of the components of the agent model
MITIP2006, 11-12 September, Budapest
Figure 2 gives greater detail on how the conceptual models for the agents are built, including the
market place, the OEM and the Engine agents. Figure 2 shows that each agent is relatively simple,
despite the overall complexity of the model and the questions to be answered.
Figure 2. Features of major model agents
The engine is included as an agent because it is the driver for revenue and cash flows between
partners, both through initial sales and through the accumulation of engine flight hours (EFHs) and
engine flight cycles (EFCs), as well as maintenance and consumables whilst in service. The engine
is a replicated agent, which means that there are multiple instances of an engine in the model at
any point in time. Each of these engine instances exhibits individual behaviour dependent upon
external factors. For example, in this model, engine usage has a variety of characteristics (e.g.
flying hard/flying easy, short/long haul, preventative maintenance/maintain when fails) dependent
on the attitude, the business model and the approach of the customer.
The model runs throughout the lifecycle of the engine programme, from initial product
development, when the product is launched, through sales, engines in use and ultimately retiring.
Activity between agents is generated by the sale of the engine, which get planned and assembled
by the OEM, leading to communication and orders being placed on the supply base. As the engine
is delivered to the customer, its flight life begins and usage is simulated. This triggers demand and
activity in the aftermarket. A simulation run therefore represents approximately 50 years. Each
agent has parameterised inputs, which create behaviours and influence decision making. All data
used within the model was provided through the project’s industrial partners.
The
simulation
model
has
been
developed
using
the
AnyLogicTM
software
TM
(http://www.xjtek.com/anylogic/). AnyLogic is an object orientated Java based simulation tool that
is ideal for agent-based logic. Integrated into the model is a SQL Server 2005 database to provide
a simplified ePlatform for inter-company communication and data sharing, and financial accounting
tools for the recording data used by each agent for decision making. These represent the
mechanism for the exchange of data seen in the real system.
5. RESULTS – AN EXAMPLE EXPERIMENT
The experimentation conducted to date has been to illustrate the potential applicability of the model
to support strategy decision making. Table 1 shows potential scenarios considered. The results
explore what this may mean for the value chain partners. The demand slump in scenario 4
captures the implications of a significant sudden disruption to the demand for new engine sales
and the usage of the existing fleet, reflecting an event similar in magnitude to ‘9/11’. The agentbased model can then be used to understand how such an event may impact on the value chain
under a range of different business strategies. Figure 3 shows the payback curves of a sudden
slump situation occurring at year 8 of the programme with a staged recovery over a 5 year period,
and the comparable non-slump simulation run. The scenarios vary from engine sales
predominantly based on the bundled product and service ‘TotalCare’ offer to engine sales
predominantly based on maintenance sold on a ‘Time & Materials’ basis.
MITIP2006, 11-12 September, Budapest
Name
Description of future business environment
1. Time and Material dominated
The model is run using a predominantly 'Time and Materials' sales environment (80% of transactions).
Aftermarket revenue based on sales of spare parts and time worked on overhaul and repair of engines.
2. Move towards 'TotalCare'
This simulation experiment represents a move from scenario 1, with 60% of sales based on TotalCare. In
TotalCare, revenue is linked to flying hours and engine cycles rather than maintenance activity.
3. TotalCare dominated
A continuation of the trend in scenario 2, with TotalCare sales now at 80% of all transactions.
4. Demand slump
A comparison between scenario 2 and 1 with a sudden slump in demand at year 8 and recovery over a 5
year period (e.g. a major event, such as a 9/11, with immediate mothballing of aircraft and suspension of
new sales. Parked engines gradual re-introduction at year 2 and sales begin to recover to previous levels.
Table 1. Examples of the experimentation conducted to date
Cash Flows
$25,000,000
$20,000,000
B a la n c e ( $ )
$15,000,000
$10,000,000
$5,000,000
$0
-$5,000,000
1
101
201
301
401
501
601
701
801
-$10,000,000
-$15,000,000
Time (w eeks)
60TotalCare_Crash
60% TotalCare sales
20%TotalCare_Crash
20%TotalCare
Figure 3. Simulated performance of a value chain partner
60% TotalCare
/ 40% ‘time and
materials’
Description
60% TotalCare.
Demand crash
year 8
20% TotalCare
/ 80% ‘time and
materials’
20% TotalCare. Demand
crash year 8
Break even point (week)
466
481
662
717
Break even point (year)
8.96
9.25
12.73
13.79
NPV at 10 years
-$4.34 million
-$4.75 million
-$7.6 million
-$7.86 million
NPV at 20 years
$2.41 million
$1.65 million
-$3.86 million
-$4.75 million
14.00
13.00
7.00
6.00
IRR (20 year analysis)
Table 2. Payback analysis for a value chain partner
In both the ‘Time and Materials’ and the ‘TotalCare’ experiments the crash delays the breakeven
point. However, there is variability between experiments in the duration of the effect – in the Time
and Material simulation experiment, the delay is 55 weeks (an additional 8%), and for TotalCare
this is only 15 weeks (3%). Therefore, the strategy of adopting ‘TotalCare’ may create a revenue
system that is more robust and resilient to the impact of dramatic changes in demand. This
demonstrates how the model can be used to inform decision makers about the long term
consequences of their actions.
6. CONCLUSIONS
By its nature it is difficult to plan and evaluate business strategies. Scenario building techniques
can assist managers in conceptualising possibilities, but will not provide a clear understanding of
what the implications for the business may be over extended timescales.
The agent-based modelling tool described shows that aggregated performance for a system can
be modelled as the result of micro level interactions. The agents may be relatively simple, but their
interactions lead to complex dynamics that may be difficult to capture, represent and evaluate
using other approaches. Agent-based models can be used to understand how strategies applied at
each level of the value chain may work over a long lifecycle. Initial results have shown that the
approach has value in demonstrating the consequences of strategy decisions.
MITIP2006, 11-12 September, Budapest
Future work will show the flexibility of the simulation model in exploring and interpreting strategy
decisions. However, strategy decisions have many operational and tactical implications and the
model has the potential to consider many of these. Table 3 outlines some of strategic and
operational scenarios of real interest to the sector that are being investigated.
Name
The agile Value Chain
The disposable engine
World Regions
PMA Threat [1]
The Collaboration Hub
Capacity modelling
Sub-RRSPs
Hub-society [1]
Aircraft age [1]
Description of value chain scenario
Higher volume exerts more pressure on value chain members and, where problems exist, they may have a more
significant impact. This scenario is therefore more operationally focussed, considering how the operations systems and
management approaches perform for high volume manufacturing.
This scenario considers a more simply constructed engine with a short lifecycle, regular disposal and reverse logistics
and parts reclamation and less important aftermarket. The value chain operates in a high volume environment.
Structuring demand and engine characteristics based on Asian, European and North American markets.
PMA involves the copying of original parts by competitors and represents a significant threat to the industry. As
margins remain high in the aftermarket (subsidising new engine sales) competitors may be tempted to enter this market.
Currently information is shared using linear communication tools and message passing. The collaboration hub would
create a work-share environment that will enable parallel working and increase data visibility.
Examining where and how capacity can be structured to meet the requirements of the customer.
To reduce the risk for a tier 1 RRSP, agreement could be shared with tier 2 suppliers.
In a hub society, larger aircraft may be dominant, which may alter the mix of engines and spares requirement for the
industry – fewer aircraft engines will be required.
A variety of scenarios involving aircraft and engine retirement age can be modelled. With concerns for the environment
and pressures on oil resources, engines may be retired earlier or require major mid-life modifications.
Table 3. Future scenarios which could be explored
7. REFERENCES
[1] Bramham J, Er W, Farr R, MacCarthy B (2003), Preliminary description of the future aerospace business
environment, VIVACE project reference: 2.1/UNOTT/T/04002-0.1
[2] Doganis R (2005), The Airline Business, 2nd Edition, Routledge
[3] Farr R (2005), Preliminary interactive business environment simulator (VIBES) with interim outputs and
interim validation, VIVACE project reference: 2.1/UNOTT/T/04002-0.1
[4] Grant RM (1991), Contemporary Strategy Analysis, 3rd Edition, Blackwell Business
[5] Leombruni R & Matteo R (2005), Physica A: Statistical Mechanics and its Applications Vol 355, No 1, pp
103 -1 09
[6] Liehr M, Grobler A, Klein M & Milling PM (2001), Cycles in the Sky: understanding and managing
business cycles in the airline market, Systems Dynamics Review, Vol 17, No 4, pp 311 - 332
[7] Pidd M (2004), Computer simulation in management science 5th Edition, Wiley
[8] Sterman JD (2000), Business Dynamics: System Thinking and Modeling for a Complex World, McGrawHill
[9] http://en.wikipedia.org/wiki/Boeing_747
Acknowledgements
This work has been funded by the European Commission as part of the VIVACE Integrated Project (Sixth
Framework Programme contract number AIP-CT-2003-502917).
BIOGRAPHY
Bart MacCarthy is Professor of Operations at t he University of nottingham. He has research
interests across Operations Management and management Science. David Buxton and Richard
Farr are rsearch fellows on the VIVACE project (VIVACE project reference: 2.1/UNOTT/T/040020.1) looking at future business environmemnts for the aerospace sector and their manufacturing
and logistics requirements.