Lessons learned from
data science in
industry
Jeroen Linssen
Ambient Intelligence
Twente Data Meetup, Saxion edition
2019-11-07
1
Some personal data
Associate Lector Ambient Intelligence @Saxion
Artificial Intelligence
Human Media Interaction
Human-Robot Interaction
2
Main data points for this evening
Ambient Intelligence
Lectoraat
Data mining methodology
CRISP-DM
Use cases in industry
Lessons learned
Other applications
Predictive maintenance; fashion
3
The lectoraat Ambient Intelligence
Think
Act
Sense
Connected
Embedded
Systems
Applied Data
Science
Augmented Interaction
4
Applied Data Science in short
THINK
Sense Act
5
Ambient Intelligence: the people
6
Smart Industry
Sport (Health)
Safety (Areas)
Augmented
Interaction
Connected
Embedded
Systems
Applied Data
Science
29 members
10 FTE
~40 students
Technology Readiness Levels
Fundamental research → ‘into the wild’
Direct involvement with companies (SMEs)
Link between universities and company R&D
7
Industry 4.0
8Tao, F., Qi, Q., Liu, A., & Kusiak, A. (2018). Data-driven smart manufacturing.
Journal of Manufacturing Systems, 48, 157-169.
Data in smart industry
9
Business question → Acquisition → Analysis → Feedback
‘We want to do
something with data…!’
Where is the data?
10
Product & Manufacturing Process Design
• Marketing Research
• Demand Research
• Process Parameters
• Raw Material SCM
• Scheduling /Job shop
Production
• Equipment and Process Control
• Equipment Monitoring Maintenance
• Quality Control
• Energy Optimization
Sales & Service
• Marketing
• Supply chain management
• Customer service
CRISP-DM
CRoss-Industry
Standard Process
for Data Mining
Data acquisition
Sensor solutions (IoT)
Data analysis
Machine learning for predictive models
Data visualization
From data to information
11
The CRISP-DM cycle
Nice, but data acquisition…?
12
CRISP-DMME
13
Huber, S., Wiemer, H., Schneider, D., & Ihlenfeldt, S. (2019). DMME: Data mining methodology for
engineering applications–a holistic extension to the CRISP-DM model. Procedia CIRP, 79, 403-408.
Data acquisition at EuroMouldings
Case
 Study power consumption
 Energy usage per resource
Approach
 Continuously monitoring power
 Monitoring ‘idle time’
14
Analysis for Nijhuis
15
Analysis for Nijhuis
16
Overview of IoT frameworks
17
Overview of tools for data analysis
18
What’s possible and what might be (or not)?
‘We want to do something with data…’
- With which purpose?
- Which data?
- Is it available already?
- Can it be acquired?
- Is the data ‘rich enough’?
- What does preliminary inspection of the data show?
- Who will do it?
- How will it be integrated within the business process?
19
Scania Uptime Improvement
Zwolle: 200 trucks a day
500 euro/min stop time
Decreasing downtime
- Predicting failures
- Predictive maintenance
20
Passende Mode via Internet
https://youtu.be/XvOM29ZtIeU
21
Lessons (to be) learned
Work with a process model
(and adapt it)
Find the right tools for the trade
(and the available people)
Get your hands on the data
(prepare for the worst, hope for the best)
Integrate new information and new methodologies in the business process
(how?)
22
Thanks for your attention! Want more?
Free coffee on the 4th
floor of Epy Drost in
Enschede!
Send an email!
j.m.linssen@saxion.nl
Visit the website of
Ambient Intelligence!
saxion.nl/ami(Data mining)

Twente Data Meetup - Lessons learned from data science in industry (Jeroen Linssen, Ambient Intelligence, Saxion)

  • 1.
    Lessons learned from datascience in industry Jeroen Linssen Ambient Intelligence Twente Data Meetup, Saxion edition 2019-11-07 1
  • 2.
    Some personal data AssociateLector Ambient Intelligence @Saxion Artificial Intelligence Human Media Interaction Human-Robot Interaction 2
  • 3.
    Main data pointsfor this evening Ambient Intelligence Lectoraat Data mining methodology CRISP-DM Use cases in industry Lessons learned Other applications Predictive maintenance; fashion 3
  • 4.
    The lectoraat AmbientIntelligence Think Act Sense Connected Embedded Systems Applied Data Science Augmented Interaction 4
  • 5.
    Applied Data Sciencein short THINK Sense Act 5
  • 6.
    Ambient Intelligence: thepeople 6 Smart Industry Sport (Health) Safety (Areas) Augmented Interaction Connected Embedded Systems Applied Data Science 29 members 10 FTE ~40 students
  • 7.
    Technology Readiness Levels Fundamentalresearch → ‘into the wild’ Direct involvement with companies (SMEs) Link between universities and company R&D 7
  • 8.
    Industry 4.0 8Tao, F.,Qi, Q., Liu, A., & Kusiak, A. (2018). Data-driven smart manufacturing. Journal of Manufacturing Systems, 48, 157-169.
  • 9.
    Data in smartindustry 9 Business question → Acquisition → Analysis → Feedback ‘We want to do something with data…!’
  • 10.
    Where is thedata? 10 Product & Manufacturing Process Design • Marketing Research • Demand Research • Process Parameters • Raw Material SCM • Scheduling /Job shop Production • Equipment and Process Control • Equipment Monitoring Maintenance • Quality Control • Energy Optimization Sales & Service • Marketing • Supply chain management • Customer service
  • 11.
    CRISP-DM CRoss-Industry Standard Process for DataMining Data acquisition Sensor solutions (IoT) Data analysis Machine learning for predictive models Data visualization From data to information 11
  • 12.
    The CRISP-DM cycle Nice,but data acquisition…? 12
  • 13.
    CRISP-DMME 13 Huber, S., Wiemer,H., Schneider, D., & Ihlenfeldt, S. (2019). DMME: Data mining methodology for engineering applications–a holistic extension to the CRISP-DM model. Procedia CIRP, 79, 403-408.
  • 14.
    Data acquisition atEuroMouldings Case  Study power consumption  Energy usage per resource Approach  Continuously monitoring power  Monitoring ‘idle time’ 14
  • 15.
  • 16.
  • 17.
    Overview of IoTframeworks 17
  • 18.
    Overview of toolsfor data analysis 18
  • 19.
    What’s possible andwhat might be (or not)? ‘We want to do something with data…’ - With which purpose? - Which data? - Is it available already? - Can it be acquired? - Is the data ‘rich enough’? - What does preliminary inspection of the data show? - Who will do it? - How will it be integrated within the business process? 19
  • 20.
    Scania Uptime Improvement Zwolle:200 trucks a day 500 euro/min stop time Decreasing downtime - Predicting failures - Predictive maintenance 20
  • 21.
    Passende Mode viaInternet https://youtu.be/XvOM29ZtIeU 21
  • 22.
    Lessons (to be)learned Work with a process model (and adapt it) Find the right tools for the trade (and the available people) Get your hands on the data (prepare for the worst, hope for the best) Integrate new information and new methodologies in the business process (how?) 22
  • 23.
    Thanks for yourattention! Want more? Free coffee on the 4th floor of Epy Drost in Enschede! Send an email! j.m.linssen@saxion.nl Visit the website of Ambient Intelligence! saxion.nl/ami(Data mining)