Planning Issues in Intelligent Manufacturing Systems

1
Planning Issues in Intelligent Manufacturing
Systems

• Paul Valckenaers
• K.U.Leuven PMA
• Belgium
• Valckenaersp_at_acm.org

2
Overview

• Background information
• Planning tasks in manufacturing
• Facilities design, process planning,
  manufacturing control
• Intelligent manufacturing systems
• Specific research developments
• Emergent manufacturing control and stigmergy
• Planning tasks in emergent control
• Other issues and conclusions

3
Background

• Engineering _at_ K.U.Leuven
• Computer Science (S/W Appl. Maths)
• PhD in mechanical engineering (production
  automation)
• Projects HMS, Mascada, Magecc, MPA
• Agent-based manufacturing control
• Design of emergent systems
• Integration issues
• Critical mass for s/w components

4
Planning in Manufacturing (1)

• Facilities (re)design
• Requirements (capturing of)
• Building blocks (modular plants)
• Guarantee safety, cleanness
• Re-usability
• Design of manufacturing plants
• Adapt to changes, to new locations
• Assess performance of design alternatives
• Predicting performance of a design is hard
• Hinders the application of planning technology

5
Planning in Manufacturing (2)

• Process planning
• Recipes
• Avoiding resource allocation choices ?
• Information representation
• Lazy, on-demand
• Dialogues
• Validation issues
• Lock-in issues
• Technology product specifics
• Grounding issues

6
Planning in Manufacturing (3)

• Manufacturing control
• Internal plant logistics
• Routing, resource allocation, starting of
  processes
• Detailed scheduling schedule execution
• Objectives of manufacturing control
• Going concerns
• No objective function ?
• Productivity (throughput in money) first
• Smooth and effective flows // chess analogy
• Lead time, wip, due dates second
• Not all tasks are equal

7
Intelligent Manufacturing Systems

• Product agents
• Resource agents
• Order agents
• Staff agents
• Architecture
• Basic P R O

• Basic agents
• Responsible
• Reflect what is/exists
• Integration
• Staff agents
• Know-how, legacy
• Support, advise
• OO
• Specialisation
• Aggregation

8
Concerns in Manufacturing Control

• Feasibility
• Recipes, deadlock...
• Thrashing, starvation...
• Operations
• Load balancing, lead times, batching...
• Staff
• Initial solutions, guidance...

9
Tasks in Manufacturing Control

• Providing information (reflection)
• Feasibility, operational, staff
• Decision taking (fragile)
• Uses (and respects) the information
• Adaptable without causing software maintenance
  cascading to the information providing parts
• Issue stable agents (s/w components)
• Challenge global coordination ( system-wide)
• Stigmergy ants provide the inspiration

10
Stigmergy and Ants

• Stigmergy
• Indirect interaction
• Signs in the environment
• Lightweight compared to direct negotiation
• Ants foraging for food
• Simple rules, emergent coordinated behaviour
• Global information is locally available
• Environment is part of the solution

11
Emergent Load Forecasting

• Resource agents reflect resources
• Environment is part of the solution
• Full life cycle, topology, ...
• Locations have attached information spaces
• What-if mode
• Ex. Conveyor belt agent
• Attributes, observers...
• Modifiers/actions
• What-if services

12
Emergent Load Forecasting

• Attributes/observers
• Information spaces attached to ...
• Modifiers/actions
• Life cycle
• Create, delete, connect, disconnect...
• Usage
• Start, stop...
• Synchronisation with reality

13
Emergent Load Forecasting

• What-if services
• PropagateUpstream
• PropagateDownstream
• Adapt time information
• Register intentions
• GiveLoadForecast
• Based on intentions communicated by ...

14
Emergent Load Forecasting

• Order agents create mobile agents that virtually
  move across the resources
• Mobile agents behave as if the present time is
  their estimated arrival time
• Mobile agents have the resource agents forecast
  their travel times
• Mobile agents use the content of the info spaces
• Mobile agents reflect decision mechanism of the
  order agent (without making assumptions about it)

15
Emergent Load Forecasting

• Mobile agent inform resource agents about the
  order agents intentions
• Resource agents combine intentions into a load
  forecast
• Order agents and their mobile agents have a
  tendency to stick to declared intentions
• When visiting a processing unit, the product
  agent and resource agent predict processing time
  and ...
• Mobile agents may send back an estimated arrival
  times ( lead time estimations)

16
Emergent Load Forecasting

• Critical resource agents (processing units)
  propagate load forecasts upstream
• This information is copied into the information
  spaces attached to the resource agents (typically
  attached to exits, entries...)
• Upstream propagation adapts the forecasts to
  account for transportation time, ...
• Evaporation refresh

17
Emergent Load Forecasting
18
Planning in Emergent Control

• Lead time
• Throughput
• Feasibility
• Flow balancing
• In-flow control
• (Spare capacity)

• Products travellers
• Factories roads
• Right product
• Proper destinations
• Heavy traffic
• Loaded factories

19
Planning in Emergent Control

• Grouping
• Mutual influencing
• Transactions

• Heat treatment
• Airplane load
• Exit blocking
• Aircraft unloading
• Do it properly
• Or not at all

20
Planning in Emergent Control

• Forecasting
• Individual
• Intensions
• Emergent
• On-line
• Nervousness control
• Bottle-necks
• (The goal)

• Prevent problems from happening, both in
  factories and in traffic
• Handle problems on detailed customer needs
• Make forecasts possible
• Production capacity
• Road capacity

21
Non-functional Requirements

• Emergent system design to cope with complexity
  and dynamics
• Limited exposure to survive
• Critical mass relative to artefact complexity

22
Concluding Remarks

• System engineering properties
• Stigmergic design
• Limited exposure, emergence
• Global information locally available
• Evolving systems
• Actual control decisions come last
• Multi-level planning issues
• Staff option

23
Intelligent Manufacturing Systems

• Reality is always consistent
• Ironware is part of the reality for MACC
• Essential models in OO-design
• Reflecting reality - full life cycle
• Xmas lists often are inconsistent
• Xmas lists change frequently
• Functional decomposition is fragile