Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
MACHINE TOOL SENSORIZATION TO IMPROVE THE PERFORMANCES
Sensori e Data Fusion nelle Lavorazioni Meccaniche
10 Aprile 2013, Piacenza, Italy Marco Grasso Supervision: Bianca Maria Colosimo Area 2 Person in Charge: Giovanni Moroni
15/05/2013 2
High Performance Manufacturing
High Performance Manufacturing
& Intelligent Machine
Tools
Autonomy
Automated supervision
Adaptive control
Condition-based maintenance
Waste and defect reduction
High quality
High productivity
Machine to machine, etc.
Machine Tool Condition
Monitoring
Process Quality/Stability
Monitoring
What?
How?
Sensors
Diagnostics & Prognosis
Error mitigation /
suppression
Maintenance
Bearing damage
Tool breakage
Tool chipping
Vibrations
15/05/2013 3
The Intelligent Machine Tool?
Macchina utensile per lavorazione leghe metalliche
Adaptive Control & Process Optimization
Decision Making
Product specifications
Technological constraints
Productivity Goals Available vibration mitigation strategies (Spindle speed variation, Speed Tuning, Feed Tuning)
Background process knowledge
Machine Tool
1
X
Y
Z
RAM_Z_ACCIAIO
0.015122
.030243.045365
.060486.075608
.090729.105851
.120972.136094
MAR 18 2010
17:59:17
NODAL SOLUTION
STEP=1
SUB =3
FREQ=293.227
USUM (AVG)
RSYS=0
DMX =.136094
SMX =.136094
x
y
kx
cx
kycy
x
y
kx
cx
kycy
Machine - Process Model Experimental Characterization
Sen
sors
Dat
a A
cqu
isit
ion
&
Sig
nal
P
roce
ssin
g
Feat
ure
ex
trac
tio
n /
se
lect
ion
Process
Performance Monitoring
15/05/2013 4
How to improve the performances?
The first ingredient: SENSORS
1
Our everyday life depends on sensors (smart-phones, tablets, vehicles, controllers, etc.) Devices we use every day are equipped with any kind of sensor Everyone of us can be a sensor node of a global network SenseAble City – MIT Lab Google – Real-Time Traffic monitor
15/05/2013 5
How to improve the performances?
The first ingredient: SENSORS
1
What about Machine Tools? Off-line
CAD CAPP CAM
Part Shape &
Specs
On-line
Part P/G NC Code
Final product
Quality Assurance CMM
Machine Tools are sold “naked”
Process
Monitoring on Final Product
In-process
monitoring rarely performed
(external toolkits)
Adaptive control rarely used
The current situation
More effective
usage of already available sensors
Integration of
additional sensors
Data fusion
Efficient/robust signal processing
Integrated
adaptive control
Machine to Machine
Intelligent Machine
15/05/2013 6
How to improve the performances?
The first ingredient: SENSORS
1
Which type of sensors?
Low-cost sensors
Smart sensors
Non-intrusive sensors
Wireless Communications
Autonomous sensors Sensor Networks
Etc…
A single highly informative data source may be difficult to have in industrial applications. It implies high costs, high intrusivity, complex installation, … A better approach:
Distributed Data Sources
Optical sensors
15/05/2013 7
How to improve the performances?
The first ingredient: SENSORS
1
Cutting Process Operator
Is there a problem?
Intervention
In-Process Monitoring
Sensors (vibration, force,
acoustic emission, lasers, etc.)
Is there a problem?
Operator Adaptive Control
Remote supervision
15/05/2013 8
DATA FUSION
How to improve the performances?
…
SENSORS
???
0 50 100 150 200 250-2000
0
2000
4000
Forz
a X
[N
]
50 100 150 200 250-500
0
500
1000
Forz
a Z
[N
]
0 10 20 30 40 50 603100
3200
3300
3400
3500
3600
Time [s]
Pout
[bar
]
0 100 200 300 400 500 600 700 800 900 1000-100
-50
0
50
100
150
200
250
X: 223.3
Y: 110.8
Frequency (Hz)
LOG
Powe
r
SIGNALS
External Info / Background Knowledge
Sensors alone do not transform a “stupid” machine into an “intelligent” machine The second ingredient: SIGNAL ANALYSIS and DATA FUSION
2
Statistical Approaches
Machine Learning
Approaches
How to combine information from multiple sources?
15/05/2013 9
How to improve the performances?
Sensors alone do not transform a “stupid” machine into an “intelligent” machine The second ingredient: SIGNAL ANALYSIS and DATA FUSION
2
Data Fusion Combination of data made available by different sources, in order to provide a better understanding of a given process, phenomenon, system, etc.
Benefits More complete characterization of phenomena Robustness with respect to disturbances and changing working conditions More effective use of available data More efficient use of available data Higher decision making reliability Uncertainty reduction in inference processes
S
S
S
S
S
S
S
DATA FUSION
Information from single sensor: • incomplete • affected by uncertainty • dependent on working conditions • not fully reliable (sensor faults)
15/05/2013 10
How to improve the performances? Sig
nal Acquis
itio
n
Alignment
Segmentation
Common Format
Conversion
Reconstruction
Processing Elaboration
Feature Selection and Extraction
Dimensional Reduction
/ Information Synthesis
Decision Level
Fault Detection
Fault Classification
Prognosis
Corrective / Mitigation Strategy Selection
Support to Operator Decision
Making
DATA FUSION Low Level High Level
Raw Signals
New Features
Processed Signals
Interpretation, Monitoring, Diagnosis
Sensors alone do not transform a “stupid” machine into an “intelligent” machine The second ingredient: SIGNAL ANALYSIS and DATA FUSION
2
From low level to high level data fusion
11.5 12 12.5 13 13.5 14 14.5 15 15.5 16 16.57000
7500
8000
8500
9000
9500
10000
Corrente assorbita azionamento Y [A]
Ris
ultante
forz
e d
i ta
glio
[N
]
15/05/2013 11
How to improve the performances?
Fusing multiple information sources may provide a better interpretation of a phenomenon
Sensors alone do not transform a “stupid” machine into an “intelligent” machine The second ingredient: SIGNAL ANALYSIS and DATA FUSION
2
Feature 1
Feat
ure
2 Multi-
sensor Data
Fusion
Single signal processing
Feat
ure
a
0 10 20 30 40 50 60-5
0
5
10
Time
Featu
re a
Time
Bad condition Good condition
?
New tool
Broken tool
Worn tool
15/05/2013 12
Water pressure sensor
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
x 104
0
20
40
60
80
100
120
140
160
180
Data point
Cors
a c
ilin
dri [
mm
]
0 1000 2000 3000 4000 5000 6000 7000 80003100
3150
3200
3250
3300
3350
3400
3450
3500
3550
3600
Data point (tempo)
Pre
ssio
ne [
bar]
-200 -150 -100 -50 0 50
05
01
00
15
02
00
QDA - Pistone 1
PC1P
C2
Info
rmat
ion
syn
thes
is
Fault Classification
Plunger displacement sensors
Broken orifice
Cracked cylinder
Normal working condition
Cracked/worn valve
An example: Health condition monitoring of Ultra High Pressure pump in Water Jet cutting
Principal Component
Analysis
Pre
ssu
re s
ign
al [
bar
]
Dis
pla
cem
ent
[mm
] Data
Fusion
Examples of MUSP activities
Sensors alone do not transform a “stupid” machine into an “intelligent” machine The second ingredient: SIGNAL ANALYSIS and DATA FUSION
2
Time
Time
15/05/2013 15
How to improve the performances?
On-board data provide fundamental information for monitoring and adaptive control
They could allow monitoring capability without added sensors
Integration is mandatory to allow closing the loop for adaptive control
External sensors should be integrated with information already available on-board The third ingredient: ON-BOARD DATA AVAILABILITY and INTEGRATION
3
Courtesy: Siemens
• Axis position (current and command) • Position errors • Axis speed • Axis torque • Axis current & power • Spindle speed • Spindle torque • Spindle current & power • State triggers • Etc.
15/05/2013 16
How to improve the performances?
An example: usage of drive signals for tool condition monitoring (in milling)
External sensors should be integrated with information already available on-board The third ingredient: ON-BOARD DATA AVAILABILITY and INTEGRATION
3
1.114 1.116 1.118 1.12 1.122 1.124 1.126 1.128 1.13 1.132 1.134
x 104
5
10
Y a
xis
Curr
ent
[A] PROCESS D
1.114 1.116 1.118 1.12 1.122 1.124 1.126 1.128 1.13 1.132 1.134
x 104
0
50
100S
pin
dle
Curr
ent
[A]
1.114 1.116 1.118 1.12 1.122 1.124 1.126 1.128 1.13 1.132 1.134
x 104
-50
0
50
Data point
X a
xis
Curr
ent
[A]
Tooth breakage
Spindle speed, current, power, …
Axis position, current, power, …
60 70 80 90 100 110
100
102
104
Feature vector index (j)
T2 C
ontr
ol C
hart
MOVING PCA CHARTS - Process D
60 70 80 90 100 110
100
102
104
Feature vector index (j)
Q C
ontr
ol C
hart
breakage
breakage
Fault detection without external sensors!
Courtesy: Siemens
15/05/2013 17
How to improve the performances?
The recipe is not complete. Several open issues must be faced with:
4,5,…
Data Fusion
Fieldbus protocols
Devices from different vendors
Need for data accessibility
Need for open architectures
Need for standardization and interoperability
Need for extendibility and portability to and from different platforms, etc…
Software development
platforms
Interface standards
Data formats
Remote comm.
protocols
15/05/2013 Piè di pagina 18
Conclusions - the innovation waves
Source: General Electric report – November 2012
The future of industrial automation (mechatronics 2.0?)
How to react to the next wave?
15/05/2013 19
The seven zeros 1. Zero defects 2. Zero (excess) lot sizes 3. Zero setups 4. Zero breakdowns 5. Zero (excess) handling 6. Zero lead-time 7. Zero surging
Quality
Maintenance
Productivity
Reliability
Autonomy
Supervision
Performances
Usage of sensors is a fundamental step towards the Factory of the Future In-process measures will be more and more available: Vision IR Current, power, voltage Forces Torques Vibrations Displacements Sound emissions Ultrasounds Acoustic Emissions Temperature Pressure Flow Etc…
Conclusions
15/05/2013 20
What’s next?
End user experience On-field know how Backgound knowledge
Monitoring algorithms Empirical models Machine learning Data fusion, etc…
We need to achieve the industrial implementation of sensor-based tools
We need to broaden our industrial collaborations to collect that experience, know how and background knowledge END USER experience is a fundamental resource to actually move to the Smart Factory
Transfer of experience, know-how, knowledge into machine automation
and smart toolkits
Final goal:
15/05/2013 21
Related to sensorization, data fusion & process monitoring Recent and On-going projects
High Performance Manufacturing Framework: MIUR Partner: MUSP Consortium + national network
MILL4D Framework: Regione Emilia Romagna Partner: Capellini, ITIA-CNR
Michelangelo Framework: MIUR Partner: ITIA-CNR + international network
Tecnopolo Framework: Regione Emilia Romagna Partner: MUSP Consortium
Tapping Process Monitoring Framework: Direct Contract Partner: Marposs/Artis
STEMMA Framework: Regione Emilia Romagna Partner: MUSP Consortium
AcquaControl Framework: Regione Lombardia Partner: Tecnocut, Altag, Polimi,…
MuProD Framework: Factory of the Future Partner: Polimi + international network
EROD Framework: Industria 2015 Partner: Jobs, BIESSE, Polimi ,…
…