NSF Engineering Research Center for
Reconfigurable MachiningSystems
Reconfigurable Manufacturing Systems
A. Galip Ulsoy, Center Deputy DirectorWilliam Clay Ford Professor of Manufacturing
University of Michigan, College of EngineeringApril 11, 2002
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 2
Outline
• Reconfigurable Manufacturing Systems (RMS)– Basic concepts– ERC/RMS research plan
• Capacity Management Policies• Combined Design and Control• Concluding Remarks
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 3
Develop Product A
Design & BuildManufacturing System
Product Ain-market
Ramp-Up
Concept
Produce A for 10 - 20 years
Economic Goal: Shorten System
Design & Reconfiguration Lead-TimeP
rese
nt
Product design timereduced by CAD
System Lead Time
Exactly the functionality needed… … Exactly when needed
Develop Product B
Develop Product C
Reconfiguration
Produce A & B
Produce B & C
RU RUProduce A
Develop Product A
Time
Design & BuildManufacturing
System
Product Ain-market
RampUp
GOAL
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 4
Comparison of Three Types of Systems:
Capacity & Functionality
Functionality
Ca
pac
ity
ProductA
DedicatedMfg. Line
MultipleProducts
ProductB + C
Exactly the functionality and capacity needed . . .. . . Exactly when needed.
F M S
ProductA + B
R M S
R M S
R M S
A
A + B
B + C
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 5
Reconfiguration Scenario400,000
300,000
0 3 6 8
Actual
Projection A
Product A
Year
Product B
Product A; 300,000 parts/year
Year 0
A A
Product A; 200,000 parts/year
B
Line 2
Line 1A & BYears 3&6
A
Product A or B; 200,000 parts/year
Line 2
Line 1Year 8A
B
LaserStn.
A & B
Product A; 200,000 parts/year
Product A or B; 250,000 parts/year
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 6
Optical Measurement for Quick Ramp-Up
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 7
Rapid ramp-up is critical for successful reconfiguration,but it’s also useful for traditional systems
6 [mm]
8.0without SoV
(Data from a real plant)
6.0
4.0
2.0
12 24 36
10.0
with SoV
Installation Time [weeks]48
Ramp-Up Methodology
New methodology for systematic ramp-up of large systems based on stream-of-variation theory (SoV)
• Utilizes state-space modeling combined with statistical analysis methods
• Needs high-speed high-accuracy measurement
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 8
Reconfigurable?
DedicatedFlexible
Reconfigurable
• Reconfigurablity concept - wrenches
• RMS: reconfigure capacity and functionality
• Reconfigurable vs dedicated vs flexible is an economic decision
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 9
RMS Science Base
A set of theories and laws that are applicable to the synthesis and analysis of RMSs
and share key characteristics
RMS Science Base
Controllerconfigurationmethodology
Reconfigurablemachine
design theory
Networked Control
Stream-of-variation theorySystem
configuration rules f(machine reliability,
quality,…)
Life-cycle economicModeling
RMT machine diagnosis &calibration
Convertibility
Customization
Integrability Diagnosability
Modularity Scalability
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 10
Major Research Issues in RMS
Part Family
Market Demand:Volume & Mix
ReconfigurableSystemDesign
CoP 1System-Level
Design
Library ofMachine
Modules &MachinesReconfigurable
Machinesand Controls
CoP 2Machine / Control-Level Design
Ramp-Up Methodology
CoP 3Calibration and Ramp-Up
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 11
Center Projects
RMSDesign
RMSRamp-Up
ReconfigurableMachines
Part FamilyMarket ChangesVolume & Mix
S2Scalability& MaterialHandling
S1System-LevelConfigurator
& Process Plan
M1ReconfigurableMachine Design
Methodology
M4Network-Based
Control
R2Reconfigurable
ProcessMonitoring
S3System
ConfigurationImpact
M3Modular Logic
Control
M2&M5SpindlesFor RMTs
R4Machine Vision
R1Stream-of-Variations
For SystemDiagnostics
S4Life-CycleEconomicModeling
R3Rapid OpticalMeasurement
of Parts
CoP 3CoP 2CoP 1
Exploratory Projects
Testbed Projects for proof-of-concept and integration
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 12
Center Testbed Facility
A brief (1.5 minute) video presentation. Also an ERC Testbed web cam is available live at
http://erc.engin.umich.edu/webcam.htm
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 13
Delphi study involving experts from around the world:
•Reconfigurable Manufacturing Systems identified as one of six grand challenges
•Adaptable and reconfigurable systems - priority technology number 1
•http://www.nap.edu/readingroom/books/visionary/
National Research Council: Visionary Manufacturing Challenges for 2020
The impact of the ERC/RMS, in just 5 years, has been worldwide
RMS featured in CIRP 99 keynote paper, JUSFA 00 and JIMTOF 00 keynote papers, ASME conferences/publications, 1st CIRP RMS conference 01 and Feb. 02 issue of ASME ME magazine.
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 14
Outline
• Reconfigurable Manufacturing Systems (RMS)• Capacity Management Policies
– Capacity management via feedback
• Combined Design and Control• Concluding Remarks
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 15
How much?How much?
Capacity Management:
Motivation
Time
• A Reconfiguration Decision Consists of Two Parts:– When?– How Much?
Capacity
Market Demand
When?
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 16
Capacity Management ProblemOptimize Cost Function Wi(Ci, Di, yi) w.r.t:
Time
X][
},min{
1 XTiii
iii
XCC
DCy
−+ +==
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 17
Capacity Management Problem
Time
Capacity &
Market Demand Demand
(Slope=d)
• What has been done?– Capacity Management Policy– Uniform capacity expansion [Manne, 1961 & 1967]
Expansion Size (x)
€
x ≈2Ad
Br
Δt ≈2A
Bdr
Optimal Solutions:
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 18
Capacity Management - Background
Cost Function:
∑∞
=
−+=
0
)(n
d
nxr
eBxAW
Fixed Cost Variable CostExpansion Size
Discount Rate
Demand Slope
Expansion Size (x)
Cost (W) Optimum Value
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 19
Manne’s Example: Optimal Results
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 20
Manne’s Example: Results• What are the issues?
– Slope might change– Discount factor might change
• In addition to minimum cost, we need to consider sensitivity and robustness
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 21
Capacity Management via Feedback
i i
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 22
Manne’s Example: Feedback Approach
i i
Controller Design
C(t)
€
[D(t) −C(t)]dt = iα0
ti∫
α =A
Br⇒ Δt =
2A
Bdr
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 23
Manne’s Example: Results w/ Feedback
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 24
Manne’s Example: Results w/ Feedback
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 25
Stochastic Demand Example
Stochastic Market Demand [Freidenfelds, 1980 & 1981]
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 26
Stochastic Demand Example: ResultsStochastic Market Demand [Freidenfelds, 1980 & 1981]
Mean and variance of cost are both reduced using the feedback approach
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 27
Capacity Management: Optimal Policies
• Numerical Example and Results (A=R=0)
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 28
1. New Design 2. Change3. Respond4. Quick5. Time Delay
Introduce Dynamics In The Problem
Time
Capacity &
Market Demand
Capacity Management:
Effect of Delay
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 29
Capacity Management:
Problems Due to Delay
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 30
Capacity Management:
Optimal Policy Which Accounts for Delay
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 31
Outline
• Reconfigurable Manufacturing Systems (RMS)• Capacity Management Policies• Combined Design and Control
– Combined optimality conditions and coupling
• Concluding Remarks
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 32
Plant/Controller Optimization Strategies
Optimize the plant
Optimize the controller
Optimize the system by varying both plant & control
Optimize the plant without compromising control
performance
Optimize the controller
Optimize the system by varying the plant
Sequential
Iterative
Simultaneous
Nested
Optimize the controller
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 33
Sequential vs. Simultaneous Strategies• Definitions
• Theorem 1: The plant and controller optimization problems are coupled in the sense that their sequential solution does not necessarily give a combined optimum.
Vector of plant design variablesVector of controller design variables
Combined objective functionSet of feasible plant/controller designs
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 34
Combined Plant/LQR Control Optimization
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
⎟⎟⎠
⎞⎜⎜⎝
⎛+
+
∫T
tc
p
t
o
dttttLTTw
ew
)),(),(()),((
)(
min)(, uxx
d
ud φ
:.tos 0dh =)(
0dg ≤)(
),),(),(()( duxfx tttt =•
0du ≤),),(( ttη
0x =)),(( TTψoot xx =)(
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 35
Combined Optimality Conditions
0dh =)(0dg ≤)(
0d =Tβ
0≥β
βα TTe ddd gh0 ++= ( )T
T
tp
c
o
dtw
w⎟⎟
⎠
⎞
⎜⎜
⎝
⎛+− ∫ μηλ ddf
0du ≤),),(( ttη),),(),(()( duxfx tttt =
•oot xx =)(
0x =)),(( TTψ
( )( ) 0
xxx
=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
−++
−++
)(Td
dTfL
T
T
T
TTtt
λνψφ
λνψφ
0=ημT
0H uu =− μηT
0≥μ
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
⎟⎟⎠
⎞⎜⎜⎝
⎛+
+
∫T
tc
p
t
o
dttttLTTw
ew
)),(),(()),((
)(
min)(, uxx
d
ud φ
:.tos 0dh =)(0dg ≤)(
),),(),(()( duxfx tttt =•
0du ≤),),(( ttη
0x =)),(( TTψoot xx =)(
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 36
The Coupling Term/Decoupling Cond.s
• The coupling term:
• Sufficient decoupling conditions:– Pareto limit condition (special case of boundary decoupling).– Zero coupling term condition (interior decoupling).
• Necessary decoupling condition (interior & boundary decoupling):– When the contribution of the optimal attainable control
performance to the system objective cannot be enhanced without violating active plant design constraints, the two problems decouple.
( )T
T
tp
c
o
dtw
w⎟⎟
⎠
⎞
⎜⎜
⎝
⎛+− ∫ μηλ ddf
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 37
Summary and Conclusions• Reconfigurable Machining Systems (RMS)
– Can be (re)configured in response to market– Reduce lead time, including ramp up– Center research organization
• Capacity Management Policies– Capacity management problem– Optimal robust capacity reconfiguration policy based on feedback– Results (deterministic & stochastic) and effect of delay
• Combined Design of Plant and Controller– Sequential, iterative, nested and simultaneous strategies– Coupled problem formulated and existence of coupling proven– Optimality conditions and coupling
NSF Engineering Research Center for Reconfigurable Machining SystemsUniversity of Michigan College of Engineering
RMS Seminar # 38
Acknowledgements and References• Acknowledgements
– Collaborators: Farshid Asl, Hosam Fathy, Yoram Koren, Panos Papalambros
– Sponsor: NSF Grant EEC 9529125
• References– RMS:
• Koren et al, CIRP Keynote Paper 1999
– Dynamic Modeling of RMS: • Asl, Ulsoy and Koren, JUSFA 2000
• Asl, Ulsoy, ACC 2000, JUSFA 2002 & IMECE 2002
– Combined Design and Control:• Fathy, Papalambros and Ulsoy, ACC 2001 & DET 2001