Multiagent-based approach for the automation and quality assurance of the small series production

© WZL/Fraunhofer IPT

Multiagent-based approach for the automation and quality assuranceof the small series production16th IEEE International Conference on Emerging Technologies and Factory Automation – ETFA 2011September, 7th 2011University Toulouse 1 Capitole, Toulouse, France

Robert Schmitt, Tilo Pfeifer, Alberto PavimLaboratory for Machine Tools WZL, RWTH Aachen

Marcelo Stemmer, Jomi Hübner, Mario RoloffDepartment of Automation and Systems DAS, UFSC Florianópolis

Seite 2© WZL/Fraunhofer IPT

Conclusions and outlook6

Application example: Agentification of an automated inspection machine5

Agent-based approach as basis for the Cognitive Metrology concept4

Vision of the project Cognitive Metrology3

Self-optimizing systems and the concept of Cognitive Metrology2

Introduction1

Contents


Small Series Production Characterisation

Small series production (SSP):Focuses on the manufacturing of a big product variety in a short period of time, while having a low production volume (possibly unitary). Time for processing the complete batch is unknown and products usually have different complexity levels.

Challenges for the Inspection of Small Series ProductionP

rodu

ctio

n vo

lum

e

Mass production

Product variety

Boundary conditions and inspection requirements in SSP Lack of predictability about the process and product behaviour

Constant creation of quality documentation Increased setup cycles and no or just few products for rigging

processes Short time to observe and provide feedback to processes during

production Difficulties for reusing information and performing corrective

actions Lack of data for decision taking

Scientific challenge!Is it possible to maintain an economical production in small series and at the same time monitor the bigdiversity of product variants and process parameters, in order to guarantee the production quality?

The (rigid) metrology strategy used within mass production is unable to cope with such conditions.Demand for new flexible and adaptive metrology strategies.







Introduction1

Contents


Increased planning efforts canbe reduced withself-optimizing systems

Higher flexibility levels increasethe planning of the system

Flexible metrology supports the secured performance of flexibleproduction systems:optical sensors provideadequate benefits

flexibility

Reduce the dilemma between scale and scope: flexible production systems

Self-optimization for reducing production control complexity

scale

scope

planning-orientation

value-orientation

self-optimization

Flexible production requires also flexible metrology strategies

Benefits of optical sensors– Touchless, non-invasive and

non-destructive– High measurement speed

(inline, in-process)– Small encapsulation,

integration to production– Wide inspection range by

combination and data fusion


Cognitive production metrology and self-optimizing systemsConceptual definition

Self-optimizingsystems

Autonomy Cognition

Flexibilityand mutability

1. Analysis of the currentsituation

2. Determination of (new) system

objectives

3. Adaption of system behaviour to surrounding conditions

Cognitive Production MetrologyAutomatic definition and application of inspection tasks for several product variants, using different flexible and adaptive measurement and inspection systems

Focus:

Quality planning: 1) automatic and dynamic inspection plan generation2) prediction of process and product quality

Measurement systems:1) flexible integration of different measurement systems2) conception of adaptive measurement systems, combining multidimensional information acquired from different sources







Introduction1

Contents


PredictiveQuality Planning

QualityEvaluation

Sensor Integration and Data Fusion

InspectionPlanning

focus on product specificationsfocus on process specifications

Planned information:Inspection plan

Real information:Information about product and process

Use Case 2Small batch assembly of printed circuit boards (PCB)

Multi-sensor basedinspection of freeform parts (e.g. automotive headlights)

Use Case 1Expertise: optical metrology, sensor fusion

Expertise: inspection planning, artificial intelligence modules

Two different case studies to validate the conceptIntegration and cooperation between project partners







Introduction1

Contents


Multiagent hybrid control structure: Hierarchical view

Work level Robots/axes Product Tools/sensors

Task level Transport Calibration Quality Inspection

Planning level Process planning Route planning Inspection planning

Coordination level Product

plannerProductplanning

Productionplanning

Actors

Productiontasks

SensorsInfor-mationcarrier

Security level Watch-dog

Splitting the system complexity into distinct responsibility levels


Flexible Production EnvironmentAgents with individual orcollective self-optimizing

behavior

Message Transporting System

Multiagent hybrid control structure: Heterarchical view

InspectionplannerProduction

planner

Processplanner

Routeplanner

AgentManagement

System

DirectoryFacilitator

Processagent

Stationagent Quality

agent

Inspectionconfiguration

Imageacquisition

IOagent

GUIagent

Watchdog

Imageprocessing

Statisticagent

Logagent

Productagent

Calibrationagent

Towards self-optimization through combination of capabilities







Introduction1

Contents


Automated test stand at WZL in Germany Not a typical SSP (similar characteristics): diversity of product

variants with different design elements worked in parallel 100% inspection of the headlight glasses is required Different failure types (geometry, material, scratches, cracks, dirt) Automated inspection approach consists of optical stations Imperfections of mechanical system and inflexible software lead to

false rejection (improve flexibility and sensing capabilities!)

Use Case 2: Autonomous Inspection of Freeform Parts

Example: automotive headlights


Use Case 2: Autonomous Inspection of Freeform Parts Integration of visual inspection systems into the test stand

Stereo System

Visi Scratch System

Visi Wave System

Flexible MV- System

Calibration Pallet


Agent-based machine control approachUse Case 2: Autonomous Inspection of Freeform Parts

Transporting system

Processagent

Stationagent

AMSagent

DFagent

Securityagent

Planningagent

Routingagent

GUIagent

Stationagent

Inspection (IP)agent

Productagent

Createagents!

Calibration

agent

product Productarrived!

I must collect data about

myself!

I need to know who

I am!

Where is the MV

inspection?

Take me there stations 1 and 2!

Ok, deal!

Product recognition!

Ok, deal!

Ok, deal!

Product X identified!

I am product X. Inspection plan!

Inspection Y!

Inspection Z!Fusion of

result A with result B!

Final quality evaluation!

Take me to inspection Y

station 3!

Take me to inspection Z

station 4!

Inspection Y: result = A!

Inspection Z: result = B!

I am OK!Next step!

Providing inspection summary!

Presenting inspection summary!

Leave inspection

system!

I am a calibration product!

I am monitoring the system

consistency!

Qualityagent

Ok, deal!

Inspection Config.

agent

Inspection (IA)agent

1- Inspection Y2- Inspection Z3- Data fusion4- Quality evaluation

Feedback to previous and further processes!

Two moves counter clockwise through stations

1 and 2!

Fusion result = C!Inside of tolerance. You are OK!

Statistic/Logagent

I/Oagent

Createproduct agent!

I am monitoring

system inputsand outputs!

I am registering

system events and statistics!

Ok, deal!

Handle transport with station agents!

Get route with routing

agent!

Go to first MV station!







Introduction1

Contents


ConclusionsSSP complicates production and quality assurance tasks

CPM defines a new paradigm in terms of metrological and quality assurance systems within SSP

Goal: to make the SSP economically viable while flexibly guaranteeing the quality of processes and products

Development of methods, technologies and services for an efficient quality management and metrology application for flexible SSP (based on SO systems)

Implementation basis on top of a hybrid multiagent control structure– Flexible integration and autonomous control of different

hardware and software modules with reactive/intelligent behaviors

– Dynamic handling of the inspection of different product variants through a set of metrology agents

– Introduction of parallelism factors (transport and self-organization) within a serial industrial inspection machine

Self-optimizingsystems

Autonomy Cognition

Flexibilityand mutability

scale

scope

planning-orientation

value-orientation


Solder Paste

Printing

Optical Inspection

Optical Inspection

SMT Components

Insertion

Solder Paste Reflow Oven

Surface Mount Technology - SMT

PCB assembly chain at LABelectron in BrazilOutlook: Cognitive Assembly of Printed Circuit Boards


OrganizationSpecifications

product

AMSagent

DFagent

Securityagent

Statistic/Logagent

SCADAagent

Qualityagent

Planningagent

Processagent

Inspection (SPI) agent

Inspection(Stencil) agent Inspection

(AOI) agent

PCBAagent

Loaderartifact

Printerartifact

Conveyorartifact

Inserterartifact

Ovenartifact

Unloaderartifact

Productagent

PCB assembly chain at LABelectron in BrazilOutlook: Cognitive Assembly of Printed Circuit Boards


Thank You Project partners The depicted research has

been funded by the German and Brazilian Research Foundations DFG, CAPES, FINEP and CNPq as part of the BRAGECRIM collaborative research initiative.


Backup


Challenges for the inspection of small series

Tendence in production:– Innovative Products and Processes– Customisation of products– Increasing number of product variants– Shorter product life cycles and smaller

production batches

Flexible small series production– One productive line for different products (flexibility)– Short setup time with reduced production costs– Autonomy and robustness– Self-learning production system

Metrology within the small series aims at the improvement of:– Quality, Robustness, Flexibility, Autonomy

Future

Batchsize

Today

Mass production

Individualised production


Generic model of an agent Platform based on services

– Decomposition of a problem– Distribution of responsibilities– Decentralisation of the control

Flexibility for the integration of hardware and software modules

Independent of programming language and operational system

Characteristics of an agent:– Encapsulation– Autonomy– Goal oriented– Reactivity– Proactivity– Interaction– Intelligence

Behaviour

Sensing

Acting

Agentenvironment

model

goals

capabilities

information

knowledge

Environment


Comparison between programming paradigms

dynamic integration

agent-based system

monolithic system object oriented system distributed system

http://d3web.sourceforge.net/index.php?page=screenshots&lang=de


Source: FLIR

Measurement on DemandFlexible inspection of PCB assembly

Documents

Multiagent-based approach for the automation and quality assurance of the small series production