Jigsaw simulation

Post on 07-May-2015

872 views 5 download

description

Presentation given to Jigsaw Consultants network during January 2011.Gives a background of DSE Consulting, Simulation and the AnyLogic software

transcript

Technology Demonstration for JIGSAW consultants network

Business & Operations Simulation in AnyLogic

CUSTOMER INSIGHTS

"although we're not new to modelling risk, the models developed by dseConsulting have created unique insights into the market we trade in. The agent-based approach used by David means that we can apply simulation to complex areas of decision making, and we are now working hard to ensure that modelling is included at each step in the strategic planning process." ANDERS NILSSON

VP BUSINESS STRATEGIES VOLVO AERO

"the AnyLogic tool has allowed enabled me to model unique levels of complexity that would not have been possible with traditional simulation tools.

I have used SD applications to model DC flows but they are limited and modelling complex interactions become so difficult they boarder on intractable - the combined modelling approach incorporating ABM removes the need to take a helicopter view of the world and concentrate on what actually happens.”

PHILIP GREENINGCRANFIELD SCHOOL OF MANAGEMENT

dseConsulting LTD• Launched Jan 08

• Simulation Modelling Consultancy– Software sales– Training– Project management– Simulation consultancy

• UK’s leading practitioner for ABMS consultancy

• And in the quieter times…– Excel automation– Optimisation– Operations Strategy & Management

dse Case studies

• Volvo Aero– Strategic investment decision– ABMS

• Rolls-Royce– Forecast and capacity planning– Mixed ABMS / DE

• Agusta Westland– Strategic Asset Management– SD & Mixed ABMS / DE

• Tesco– ‘Space’ strategy optimisation– ABMS

• Cranfield– Eng School: Commercial partner– SoM: Development partner

• UNOTT– Comp Sci: Mentoring & Project

Management

CS2: Airline Market Economics

Volvo Aero(cSeries entry)

Economic conditions

STRATEGY LAYER

AIRLINE LAYER

Assess cost efficiency of fleet

a

OEM LAYER

Launch Airframe programme

AIRBUS

Oil Market

BOEINGBOMB-ARDIER

AIRFRAME PROJECTS

Y1NSRcSeries

NSR

737A320

aa

aa

ab

SECOND HAND MARKET

AIRCRAFT LAYER

737A320

737A320

737A320

737A320

• Motivation• Make a financially optimum decision

• Attributes• ID• Business Model (Business Aviation Grp,

Virtual, Flag Carrier)• Current fleet (+ operating costs)

• Behaviours• Review current fleet• Buy Aircraft

• Negotiate with OEMs

Airline

• Motivation• Maintain / achieve market share

• Attributes• ID• Product

• Behaviours• Decide to launch new development

programme• Respond to competition

OEM

• Motivation• (Operated by airlines)

• Attributes• FH• Running costs

• Behaviours• Fly

Aircraft

• Motivation

• Attributes• Costs (direct and running)• Performance characteristics• Timeline – expected launch date,

launch date, EiS date• Behaviours

Airframe development program

CS 1: Asset lifecycle Management

Agusta Westland(Fleet planning under alternative

strategic plans)

GEO-POLICITCAL

STRATEGY LAYER

MISSION PLANNER

SCHEDULE MISSION

POLICY

BASES (AIRCRAFT + AIRCREW)

PILOT 1

TYPE 1

TYPE 1

TYPE 2 TYPE 3TYPE 2PILOT 2

PILOT 3

• Motivation• Send aircraft on missions to meet Customer

requirement• Attributes

• Operating schedule• Behaviours

• Choose the aircraft available with the least flying hours• Manage maintenance schedule to ensure a ‘balanced’

fleets

Mission Controller

• Motivation• Complete planned and unscheduled maintenance to

ensure fleet availability• Attributes

• Capacity• Lead times

• Behaviours• Implement upgrade• Conduct planned maintenance• Conduct unplanned maintenance

Maintenance Facility

• Motivation• Governs the mission type, duration and frequency

• Attributes• Location• State

Base

• Motivation• (Operated by mission controller)

• Attributes• Type• Failures• Random Events• Flying hours• Current base• Performance characteristics

• Behaviours• Fly

Aircraft (helicopter)

• Motivation• Fly the right aircraft to meet the mission

schedule• Attributes• Training log book• Flight hour log book• Current base

• Behaviours• Train• Fly

Aircrew

ANYLOGIC

the first “Simulation development environment”

• Designed to overcome the limitations of ‘shrink wrapped software’– that Quickly run out of power and scope when faced with large and

complex applications• You can mix approaches meaning that analyst can focus on the nature of the

problem and not the technique they know

– Only partially driven by point click using pre-defined objects

• Object Orientation programming language– Based in the eclipse DE– Completely scalable and can be integrated into traditional business IT

systems

• Pricing– Professional C. £12k– Standard C. £4.5k

“Simulation: the key to business success” Gartner Research 2010

ArenaExtendSimul8AutoModPROMODELEnterprise DynamicsFlexSimeMPlant…

MATLABVisSimLabViewEasy 5…

[Academic software:]SwarmRePast

VenSimPowerSimiThinkModelMaker

SD DE AB DS

The AnyLogic approach

AnyLogic – Multi-Paradigm Simulation Tool

SD DE AB DS

• You can easily vary and adjust the level of abstraction• Models are very easy to re-use• You can mix approaches• You can switch from one approach to another

Why use simulation?• As an insurance policy

– for large capital investment projects to ensure they work as envisaged– predicting realistic KPI’s of future system and refining plans and strategies

• Testing alternative designs – when testing in real world is too expensive– when testing in real world is too time consuming– when testing in real world is not possible (Non-existing infrastructure)– when testing in real world is too dangerous

• Decision facilitator– when decision makers need an unbiased model to show the impact of

decisions– when gut-feel is not enough to understand full complexity– when nobody has the expertise covering all aspects of the system– when the impact of decisions need to be communicated to multiple parties

DISCRETE EVENT SIMULATION

How can my factory double capacity?

• Entities moving through a flowchart based system, • Queuing theory based on

• arrival rates, delays and resource utilisation#

• Basic structure applicable to 100s of situations• Banks, Retail, Ports, Trains, Passengers, Logistics, operations,..

Example from AnyLogic

Strategic Questions where DE alone cannot help

• How do people interact with the system?– Why is the arrival rate X?– Why does it take X for my resource to process the order?

• What are consumption trends and how will these affect demand?

• What will the competitors response be?

• How will connected systems be affected?– Most problems are comprised of a system of systems..

SYSTEM DYNAMICS

Intrinsic to Systems Thinking

“Sales are poor (effect)…because sales staff are not highly motivated (cause & effect)…because salaries are low (cause)”

• Whole systems modelling– A structured approach to modelling a complex world– Causal loop diagrams & Feedback Structure– Systems that change continuously over time using the principle of stock and flows– Highly aggregated, highly abstract representations

• Modelling tends to be explanatory– Based on a mental model– Stock and flow can be difficult to put together and data availability is always a problem

• Some very famous examples that illustrating reasons behind a trend– Beer Game, Bass-Diffusion, “Limits to growth” (Club of Rome)

Base Diffusion model

PotentialAdopters Adopters

AdoptionRate

Adoptionfrom

Advertising

Adoptionfrom Wordof Mouth

TotalPopulation

AdoptionFraction

ContactRate

AdvertisingEffectiveness

+

+ ++

-

+ +

+

+

B

B

R

In AnyLogic this looks like…

Adoptionfrom

Advertising

Adoptionfrom Wordof Mouth

TotalPopulation

AdoptionFraction

ContactRate

AdvertisingEffectiveness

PotentialAdopters

AdoptionRate Adopters

Type: Stock Variabled(…)/dt = : Adoption_RateInitial value: 0

Type: Stock Variabled(…)/dt = : Adoption_RateInitial value: 0

Type: ParameterInitial value : 10,000Type: ParameterInitial value : 10,000

Type: ParameterInitial value : 0.015Type: ParameterInitial value : 0.015

Type: ParameterInitial value : 0.015Type: ParameterInitial value : 0.015

Type: Stock Variabled(…)/dt = : - Adoption_RateInitial value: 10,000

Type: Stock Variabled(…)/dt = : - Adoption_RateInitial value: 10,000

Type: ParameterInitial value : 0.015Type: ParameterInitial value : 0.015

Type: Flow Aux Variable… = : Potential Adopters * Advertising Effectiveness

Type: Flow Aux Variable… = : Potential Adopters * Advertising Effectiveness

Type: Flow Aux Variable… = : Contact Rate * Adoption Fraction * Potential Adopters * Adopters / Total Population

Type: Flow Aux Variable… = : Contact Rate * Adoption Fraction * Potential Adopters * Adopters / Total Population

Type: Flow Aux Variable… = : Adoption from Advertising + Adoption from Word of MouthType: Flow Aux Variable… = : Adoption from Advertising + Adoption from Word of Mouth

AGENT-BASED MODELLING & SIMULATION

“Macro from Micro”• Used where

– Large populations of things– Human centric or human influenced systems

• Agents…– autonomous – decision makers– interacting with the local environment– interacting with each other– live in a dynamic environment

• Its ‘bottom up’ modelling– agents, decisions, states

• Examples– Inter firm competition– Strategy & market dynamics

When to use ABMS..1. Your markets are fragmented and there are multiple groups or segments

within a population that may behave differently2. The past is no prediction of the future, but surprises can often be explained

with the power of hindsight3. Your problem is complex and it is difficult to predict how the system will evolve4. Any element of strategic behaviour or co-operation exists5. Learning or behaviour is important6. Your population has a natural representation of an agent and system

performance is significantly related to this agent7. Estimations of inputs are missing, unavailable or not sufficiently realistic8. Structural elements are the result, rather than an input to, your model9. There is a geo-spatial element to your model10. You are uncomfortable with the huge assumptions required by Discrete Event

Simulation (DES), System Dynamics (SD) or spreadsheet based models

* With thanks to Charles Macal whose thoughts during 2010’s OR conference inspired this list.

Key Agent building blocks - States

States can govern Deciding between strategy

Bass diffusion agent based

PotentialAdopter

Adopter

Rate:AdEffectiveness

“Buy!”Guard: randomTrue(AdoptionFraction)

Rate: ContactRate

<random agent>.”Buy!”

Doing more with ABMS

Opportunities to find out more…?

• dave@dseconsulting.co.uk• www.dseconsulting.co.uk• http://uk.linkedin.com/in/dseconsulting• 0121 288 0503

• Training options1. Cranfield short course programme: 2 day short course covering all

three techniques

2. 3 days official AnyLogic training (March 2011, London)

3. OR Society training: 2 day course covering ABMS only in collaboration with UNOTT (December 2011, Birmingham)