Predictive Maintenance with R
•About eoda 
•Predictive Maintenance 
•Predictive Maintenance with R 
•Results as a Service 
Agenda
About eoda 
•an interdisciplinary team of data scientists, engineers, economists and social scientists, 
•founded 2010 in Kassel (Germany), 
•specialized in analyzing structured and unstructured data, 
•integrated portfolio for solving analytical problems, 
•with a focus on „R“.
Consulting 
Software 
Solution 
Training 
eoda portfolio
Predictive Maintenance
The requirements on maintenance 
International competition 
Shorter product life cycles 
Faster technological leaps 
More complex business processes 
Shift from product to service
Evolution of Maintenance Concepts 
Reactive or Breakdown Maintenance 
Preventive or Periodic Maintenance 
Condition-based Maintenance 
Unplanned production shutdowns 
Inefficient use of resources 
Simple rules 
 Not very precise
Predictive Maintenance as an extension of condition-based maintenance represents the informatization of production processes. With intelligent IT-based production systems Predictive Maintenance represents one important step on the path towards the development of a Smart Factory in industrial production. 
Predictive Maintenance 
The future of maintenance
Predictive Maintenance 
Example – Gearbox Bearing damage in wind farm 
•Reactive Maintenance 
•Cost for a replacement of the bearing $ 250.000 
•Cran costs $ 150.000 
•Power generation / Revenue losses $ 26.000 
$ 426.000 
Source: http://www.wwindea.org/
Predictive Maintenance 
Example – Gearbox Bearing damage in wind farm 
•Predictive Maintenance 
Use of acceleration sensors, oil particle counters and weather forecast modules, plus reliable evaluation of the data 
Early detection of the damage at the gearbox bearing 
•Repair instead of exchange of the bearing $ 30.000 < $ 250.000 
•Lower cran costs $ 75.000 < $ 150.000 
•Power generation / Revenue losses $ 2.000 < $ 26.000 
$ 107.000 < $ 426.000 
Source: http://www.wwindea.org/
Predictive Maintenance 
Potential factors 
50 % Reduction of maintenance costs 
50 % Reduction of machine damage 
50 % Reduction of machine downtime 
20 % Increase in machine lifetime 
20 % Increase in productivity 
25 % - 60% Profit growth 
Source: Barber, Steve & Goldbeck, P.: “Die Vorteile einer vorwärtsgerichteten Handlungsweise mit vorbeugenden und vorausschauenden Wartungstools und –strategien – konkrete Beispiele und Fallstudien.”
Predictive Maintenance 
Time 
Data collection 
Data management 
Data analysis 
Planning of maintenance 
Maintenance 
Business Value 
Workflow
Predictive Maintenance Data Collection and Management 
Environmental Data 
Sensor-based Machine Data 
Production indicators 
Different types of data
Predictive Maintenance 
Data analysis 
Datascience 
know-how 
Requirements of the market 
Domain 
Expertise
Predictive Maintenance 
Data analysis 
Source: David Smith 
Data Scientists 
Power User 
Business User 
Service People 
Different user types with different comepetence level
Predictive Maintenance with R
Predictive Maintenance with R 
Advantages 
•Features 
•The features that come with R (without additional investment) are incomparable 
•R in the software stack 
•R can be integrated into all the layers of an analysis or reporting architecture
Predictive Maintenance with R 
Advantages 
•Features 
•The features that come with R (without additional investment) are incomparable 
•R in the software stack 
•R can be integrated into all the layers of an analysis or reporting architecture 
C 
Prototyping 
Implementation 
R directly on the machine
Predictive Maintenance with R 
Advantages 
•Features 
•The features that come with R (without additional investment) are incomparable 
•R in the software stack 
•R can be integrated into all the layers of an analysis or reporting architecture 
•Investment protection 
•The involvement of the scientific community and large companies support the development and acceptance of R 
•Quality 
•R offers high reliability and uses the latest statistical methods 
•Costs 
•R is Open Source and there are no license costs
Data Collection and Management 
Environmental Data 
Sensor-based Machine Data 
Production indicators 
Example of use: Different types of data at different times 
Predictive Maintenance with R 
Time Density 
7:30 15,3 
8:30 16,1 
9:30 15,7 
10:30 15,5 
11:30 16,0 
12:30 15,9 
Time Pressure 
7:00 235 
8:00 239 
9:00 240 
10:00 228 
11:00 231 
12:00 233
Data Collection and Management 
Environmental Data 
Sensor-based Machine Data 
Production indicators 
Predictive Maintenance with R 
Time Density 
7:30 15,3 
8:30 16,1 
9:30 15,7 
10:30 15,5 
11:30 16,0 
12:30 15,9 
Time Pressure 
7:00 235 
8:00 239 
9:00 240 
10:00 228 
11:00 231 
12:00 233 
Big Data Model based 
Density interpolation 
15,4 
16,0 
15,7 
15,4 
15,8 
16,1 
Example of use: Different types of data at different times
Data analysis 
Source: David Smith 
Data Scientists 
Power User 
Business User 
Service People 
Predictive Maintenance with R 
The comeptence level disappear with R
Predictive Maintenance with R 
Results as a Service
Data 
Analysis 
Web based Front End 
Predictive Maintenance with R 
Results as a Service eoda Service Platform 
API 
Interactive Web App 
R- Scripts 
… 
Administration 
Authentication (LDAP) 
User-, Role- Management 
Session Management 
… 
Public data sources 
Internal data 
Machine data 
Java Script
eoda GmbH Ludwig-Erhard-Straße 8 34131 Kassel Germany +49 (0) 561/202724-40 www.eoda.de http://blog.eoda.de https://service.eoda.de/ http://twitter.com/datennutzen https://www.facebook.com/datenwissennutzen info@eoda.de 
Thank you for your attention 
For more information Whitepaper: Predictive Maintenance with R www.eoda.de Results as a Service eoda Service Platform https://service.eoda.de/

Predictive Maintenance with R

  • 1.
  • 2.
    •About eoda •PredictiveMaintenance •Predictive Maintenance with R •Results as a Service Agenda
  • 3.
    About eoda •aninterdisciplinary team of data scientists, engineers, economists and social scientists, •founded 2010 in Kassel (Germany), •specialized in analyzing structured and unstructured data, •integrated portfolio for solving analytical problems, •with a focus on „R“.
  • 4.
    Consulting Software Solution Training eoda portfolio
  • 5.
  • 6.
    The requirements onmaintenance International competition Shorter product life cycles Faster technological leaps More complex business processes Shift from product to service
  • 7.
    Evolution of MaintenanceConcepts Reactive or Breakdown Maintenance Preventive or Periodic Maintenance Condition-based Maintenance Unplanned production shutdowns Inefficient use of resources Simple rules  Not very precise
  • 8.
    Predictive Maintenance asan extension of condition-based maintenance represents the informatization of production processes. With intelligent IT-based production systems Predictive Maintenance represents one important step on the path towards the development of a Smart Factory in industrial production. Predictive Maintenance The future of maintenance
  • 9.
    Predictive Maintenance Example– Gearbox Bearing damage in wind farm •Reactive Maintenance •Cost for a replacement of the bearing $ 250.000 •Cran costs $ 150.000 •Power generation / Revenue losses $ 26.000 $ 426.000 Source: http://www.wwindea.org/
  • 10.
    Predictive Maintenance Example– Gearbox Bearing damage in wind farm •Predictive Maintenance Use of acceleration sensors, oil particle counters and weather forecast modules, plus reliable evaluation of the data Early detection of the damage at the gearbox bearing •Repair instead of exchange of the bearing $ 30.000 < $ 250.000 •Lower cran costs $ 75.000 < $ 150.000 •Power generation / Revenue losses $ 2.000 < $ 26.000 $ 107.000 < $ 426.000 Source: http://www.wwindea.org/
  • 11.
    Predictive Maintenance Potentialfactors 50 % Reduction of maintenance costs 50 % Reduction of machine damage 50 % Reduction of machine downtime 20 % Increase in machine lifetime 20 % Increase in productivity 25 % - 60% Profit growth Source: Barber, Steve & Goldbeck, P.: “Die Vorteile einer vorwärtsgerichteten Handlungsweise mit vorbeugenden und vorausschauenden Wartungstools und –strategien – konkrete Beispiele und Fallstudien.”
  • 12.
    Predictive Maintenance Time Data collection Data management Data analysis Planning of maintenance Maintenance Business Value Workflow
  • 13.
    Predictive Maintenance DataCollection and Management Environmental Data Sensor-based Machine Data Production indicators Different types of data
  • 14.
    Predictive Maintenance Dataanalysis Datascience know-how Requirements of the market Domain Expertise
  • 15.
    Predictive Maintenance Dataanalysis Source: David Smith Data Scientists Power User Business User Service People Different user types with different comepetence level
  • 16.
  • 17.
    Predictive Maintenance withR Advantages •Features •The features that come with R (without additional investment) are incomparable •R in the software stack •R can be integrated into all the layers of an analysis or reporting architecture
  • 18.
    Predictive Maintenance withR Advantages •Features •The features that come with R (without additional investment) are incomparable •R in the software stack •R can be integrated into all the layers of an analysis or reporting architecture C Prototyping Implementation R directly on the machine
  • 19.
    Predictive Maintenance withR Advantages •Features •The features that come with R (without additional investment) are incomparable •R in the software stack •R can be integrated into all the layers of an analysis or reporting architecture •Investment protection •The involvement of the scientific community and large companies support the development and acceptance of R •Quality •R offers high reliability and uses the latest statistical methods •Costs •R is Open Source and there are no license costs
  • 20.
    Data Collection andManagement Environmental Data Sensor-based Machine Data Production indicators Example of use: Different types of data at different times Predictive Maintenance with R Time Density 7:30 15,3 8:30 16,1 9:30 15,7 10:30 15,5 11:30 16,0 12:30 15,9 Time Pressure 7:00 235 8:00 239 9:00 240 10:00 228 11:00 231 12:00 233
  • 21.
    Data Collection andManagement Environmental Data Sensor-based Machine Data Production indicators Predictive Maintenance with R Time Density 7:30 15,3 8:30 16,1 9:30 15,7 10:30 15,5 11:30 16,0 12:30 15,9 Time Pressure 7:00 235 8:00 239 9:00 240 10:00 228 11:00 231 12:00 233 Big Data Model based Density interpolation 15,4 16,0 15,7 15,4 15,8 16,1 Example of use: Different types of data at different times
  • 22.
    Data analysis Source:David Smith Data Scientists Power User Business User Service People Predictive Maintenance with R The comeptence level disappear with R
  • 23.
    Predictive Maintenance withR Results as a Service
  • 24.
    Data Analysis Webbased Front End Predictive Maintenance with R Results as a Service eoda Service Platform API Interactive Web App R- Scripts … Administration Authentication (LDAP) User-, Role- Management Session Management … Public data sources Internal data Machine data Java Script
  • 25.
    eoda GmbH Ludwig-Erhard-Straße8 34131 Kassel Germany +49 (0) 561/202724-40 www.eoda.de http://blog.eoda.de https://service.eoda.de/ http://twitter.com/datennutzen https://www.facebook.com/datenwissennutzen info@eoda.de Thank you for your attention For more information Whitepaper: Predictive Maintenance with R www.eoda.de Results as a Service eoda Service Platform https://service.eoda.de/