Predictive Quality in Electronics Manufacturing - Industrial Data ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Predictive Quality in Electronics Manufacturing Unrestricted © Siemens AG 2019 www.siemens.com Unrestricted © Siemens AG 2019
Siemens corporate structure
Managing Board
Corporate Services Regions Corporate Core
Power Energy Building Mobility Digital Process Siemens Siemens Financial
and Gas Management Technologies Factory Industries Healthineers Gamesa Services
Divisions
and Drives Renewable
Energy
Power
Generation
Services
Factory Control Product Lifecycle Motion Customer
Automation Products Management Control Services
Unrestricted © Siemens AG 2019
Page 2 Dr. Jochen Bönig | DF FA MF OPS&C
Excellence in manufacturing – For our customers
Our Mission Our Locations
„
Be the Role Model for Excellence in Manufacturing
to provide proven Value Add for our Customers and
„
Business Unit, based on the methods of Digital
Enterprise and Lean Industrial Engineering
Shape the Digital Future. Together.
(Karlsruhe)
Dr. Gunter Beitinger
Dr. Gunter Beitinger, Vice President (Lebanon)
Manufacturing VP of Manufacturing Amberg
Fürth Chengdu
(Nanjing)
Unrestricted © Siemens AG 2019
Page 3 Dr. Jochen Bönig | DF FA MF OPS&C
Advanced data analytics as a key enabler to optimize our
processes
How can we make
something happen?
Prescriptive
Analytics
What will happen?
„Predictive Analytics
uses Big Data to Predictive
analyze past patterns Why did it happen? Analytics Act
and predict the future“ Value What happened?
Diagnostic
Gartner Descriptice Analytics
Analytics
Analyze
Inform
Effort
Unrestricted © Siemens AG 2019
Page 4 Dr. Jochen Bönig | DF FA MF OPS&C
CRISP-DM1 as the preferred approach for our
advanced data analytics projects
Business/
Data
Process
Understanding
Understanding
Data
To get the best results Preparation
make sure to include
Deployment
process domain know-
Big Data
how! Modeling
Evaluation
1 CRISP-DM: Cross Industry Standard Process for Data Mining
Unrestricted © Siemens AG 2019
Page 5 Dr. Jochen Bönig | DF FA MF OPS&C
„Predictive Analytics requires both domain and scientific know how!“
A digital factory needs an analytics eco system
Training data Data lake
Continuous Model
Development Algorithm performance
Process refeed
Enrich
Enrich
Deploy
Streaming data Real time model execution
Algorithm results
Unrestricted © Siemens AG 2019
Page 7 Dr. Jochen Bönig | DF FA MF OPS&C
„One of the biggest challenges is to build up an analytics eco system!“
Actual EWA IT-architecture already covers a large portion of
desired Lean Digital Factory blueprint
Management
BI-Software, Manufacturing
Analytics & AI Data Platform
Product Lifecycle Management and ERP
NX/ Teamcenter Teamcenter ERP
Tecnomatix Manufacturing
~460 Clients ~70 Clients
Operations
MindSphere
Edge App
Edge App
SIMATIC IT SIMATIC Energy Scout QM SIMATIC
WinCC Analytics Edge
~1.300 Clients ~100 Clients ~1.000 Clients 16 Clients
Control
rol
SIMATIC SIMATIC SIMOTION SINUMERIK SIMATIC SCALANCE
Controller HMI IPC
20 Clients ~650 Clients ~250 Clients
Field ~2.800 Clients
SIMATIC SINAMICS SIMATIC SITOP SIRIUS
Distributed I/O Ident
Unrestricted © Siemens AG 2019
Page 9 Dr. Jochen Bönig | DF FA MF OPS&C
An edge-based system combines the benefits of a pure local
and pure cloud solution
Edge solution Cloud solution Edge-Cloud solution
Cloud
Data preparation Cloud Cockpit Model Dev
Cloud Cockpit
+ Model Dev
Feedback Compressed New
data models
Edge
Local Cockpit Data preparation Local Cockpit Data praparation
+ Model Dev + Model Runtime
Feedback
Field Feedback
Raw data Raw data Raw data
upload upload upload
Data connector Data connector Data connector
Production asset Production asset Production asset
1 Client 15 Clients
Unrestricted © Siemens AG 2019
Page 10 Dr. Jochen Bönig | DF FA MF OPS&CHot topics for advanced data analytics in manufacturing
environment
Manufacturing
“How to improve quality “How to increase uptime “How to improve
level while reducing test while minimizing collaboration while
effort?” maintenance costs?” reducing communication?”
“How to achieve one-piece- “How to stabilize
flow while maximizing processes while lowering
utilization?” control effort?”
Supplier
collaboration
Unrestricted © Siemens AG 2019
Page 11 Dr. Jochen Bönig | DF FA MF OPS&CThree use cases for advanced data analytics in
manufacturing
Manufacturing
“How to improve our quality “How to achieve one-piece- “How to increase machine
level while reducing our test flow while increasing uptime while reducing
effort?” machine utilization?” maintenance costs?”
Unrestricted © Siemens AG 2019
Page 12 Dr. Jochen Bönig | DF FA MF OPS&CAdvanced data analytics reduces test effort significantly
Today:
X-Ray
SMT 100% X-Ray
(test step)
SMT
Future:
SMT < 50% X-Ray
+
Objective: Connector PCB1 of a Problem: X-Ray test is a time and Target: Evaluate quality prediction
distributed I/O cost intensive work step model to reduce X-Ray test effort
1 PCB: Printed Circuit Board
Unrestricted © Siemens AG 2019
Page 13 Dr. Jochen Bönig | DF FA MF OPS&CInspection data and X-Ray test data are used to create a
prediction model for product quality
X-Ray test protocols
10
Solder paste inspection 01
01
11
00
10 10
01
01
11
00
10
Offline
Reflow soldering X-Ray Testing
SMD placement
Soldering paste print
Model inputs: 52Mio datasets with very high Process Quality of 7dpm1
Model requirements: Optimize test slip (Bad products are labled as good)
1 dpm: Defects per million
Unrestricted © Siemens AG 2019
Page 14 Dr. Jochen Bönig | DF FA MF OPS&CX-Ray testing effort can be reduced from 100%
to at least 70%
Model pseudo-
Next steps error
• Further data True result
model validation Slip-optimized model
n.i.O. i.O.
• Adapt algorithm to
n.i.O. 135 18373464
new product
Prediction
i.O.
• Implement 0 8378038
automatic data Class Recall 1.0 31,32%
collection and
prediction model Slip less x-Ray tests
needed!
Unrestricted © Siemens AG 2019
Page 15 Dr. Jochen Bönig | DF FA MF OPS&CMost common error types in base unit production detected
by X-Ray testing
Part orientation X1 18%
Shield pin not soldered X1 17%
Shield pin not soldered X2 14%
Defective part manufacturer X2 13%
Signal pin not soldered X1 10% X-Ray of X1 Connector
Part orientation manufacturer X2 5%
Signal pin not soldered X2 5%
Part orientation X2 4%
Defective solder joint signal pin X1 3%
Solder brige X1 3%
Unrestricted © Siemens AG 2019
Page 16 Dr. Jochen Bönig | DF FA MF OPS&CThe X-Ray system setup according to our analytics
eco-system
Model Enrich
Enrich
development
Algorithm performance
Deploy
Solder paste
Real time model
inspection data
execution
Defect?
yes
X-Ray Testing
No defect
No testing!
SMT process
Unrestricted © Siemens AG 2019
Page 17 Dr. Jochen Bönig | DF FA MF OPS&CThe panel for SMT production consists of 48 boards and is
double-sided assembled
48 47 46 45 44 43 42 41 Bottom – connector X2
40 39 38 37 36 35 34 33 48 Boards per panel
32 31 30 29 28 27 26 25 52 Pins per board_X2
2496 Pins per panel_X2
24 23 22 21 20 19 18 17
8 Field of views
16 15 14 13 12 11 10 09
08 07 06 05 04 03 02 01
41 42 43 44 45 46 47 48 Top – connector X1
33 34 35 36 37 38 3 40 48 Boards per panel
25 26 27 28 29 30 31 32 79 Pins per board_X1
3792 Pins per panel_X1
17 18 19 20 21 22 23 24
8 Field of views
09 10 11 12 13 14 15 16
01 02 03 04 05 06 07 08
Unrestricted © Siemens AG 2019
Page 18 Dr. Jochen Bönig | DF FA MF OPS&CAlternative routing of the printed circuit board depending on
the label of the algorithm
Scenario 1) All boards of the panel are predicted as good quality
Scenario 2) All boards of the panel are predicted as poor quality
Scenario 3) Some boards of the panel are predicted as good / poor quality
Edge Cloud
X-Ray Non-time-critical level
Time-critical level
(Offline)
Algorithm training
10
10 01 10
01 10
01 01 01
01 11 01
11 01
11 01 11
01 11 01
11 01
11 01 11
01 00 01
00 01
00 00
10
01
10 01
01 Time series 11
01 01
11 data 11
01
00
optimized
algorithm
Unrestricted © Siemens AG 2019
Page 19 Dr. Jochen Bönig | DF FA MF OPS&CDynamic X-Ray testing enables additional productivity
20% 80%
Process time
Handling Testing time
100
80
%
60
40
20
Handling Testing time FOV skip 0
Handling Testing
Unit
forecast
Process
time
Time/FOV € /a € /a € /a
Costs/FOV
€ /a
Unrestricted © Siemens AG 2019
Page 20 Dr. Jochen Bönig | DF FA MF OPS&CAI minimizes While maintaining the Resulting in reduced
necessary X-RAY high Quality rate of capital invest for
tests by currently 30% further X-RAY
machines of
Target for the future is
100% 100% 500k€Thank you for your attention Unrestricted © Siemens AG 2019
You can also read
