www.mayato.com Location: Düsseldorf Date: 2-3 February, 2017 Speaker: Eric Ecker, Head of Industry Analytics 1 Predictive Analytics World Manufacturing „Four steps to reliable predictions“
www.mayato.com
Location: DüsseldorfDate: 2-3 February, 2017Speaker: Eric Ecker, Head of Industry Analytics
1
Predictive Analytics World Manufacturing„Four steps to reliable predictions“
www.mayato.com
Introduction
What‘s happening out there?
Four steps to reliable predictions
Real industry cases
Agenda
2
www.mayato.com
The short formula sounds easy:
1. Collect2. Structure3. Analyze4. Visualize
Nevertheless, real life projects in various industries show that the first two steps aren‘t as trivial as they may seem. Optimized processes in automotive, efficient quality checks in steel production or predictive maintenance in medical engineering – the integration and quality of all relevant data is key to run successful predictive analytics scenarios. Furthermore, you need the right methodology to analyze the data.
Eric Ecker, Head of Industry Analytics, mayato GmbH
Industry Analytics: Four steps to reliable predictions
3
www.mayato.com
mayatoOur mission
Industry AnalyticsCustomer AnalyticsFinancial AnalyticsSecurity Analytics
We help our clients to derive value from data“Watch our predictions come true!”
www.mayato.com
What is happening out there?
5
Machines• Steam engine• Mechanizing
Mass production• Electrics• Assembly lines
Microelectronics• Computer• Automation
Data• Internet of Things• Artificial Intelligence
Industrie 4.0/IoT Agents Sensors Mobility Self-Optimization
Localization StreamingIndividual
production and configuration
Interconnectedness Intelligent Algorithms
www.mayato.com
Machines and products are getting intelligent through the usage of micro sensors, RFID tags and smallest embedded computers
These systems are self organizing, interconnected and communicate across open and standardized Internet protocols
Vast amounts of data about the state of machines, their behavior and their usage are being generated
IoT and “Industrie 4.0” provide lots of opportunities
6
www.mayato.com
Wir holen die Antwortenda raus !
015661565
5666544
568
456845664568
56456
68456
12489
245
456856
456687015505
01566
Theanswers can be found in your DATA!
www.mayato.com
And there is much more in your data …
8
Higher quality in
your production
Spare part optimization
Machinefailure
prediction
Integration of data from shop to top
floor
Higher customer
satisfaction
Less scrap
Defect analysis
Optimizationof machine
usage
www.mayato.com
Central collection of all machine and log data
Data Transformation and Cleansing
Analytics based on Data Mining techniques• Clustering• Association analysis• Classification• Anomaly detection
Visualization of results
Solution in four steps
9
www.mayato.com
Analytics projects are agile and iterative by nature
Waterfall models are more or less unsuitable
There are several iterative models like CRISP-DM
Lots of communication, discussions and decision making is needed
Analytics projects: Some advice
11
www.mayato.com
Location: DüsseldorfDate: 2-3 February, 2017Speaker: Eric Ecker, Head of Industry Analytics
12
Case Studies
www.mayato.com
Industry AnalyticsIntegrated Management Platform Ensures Production Quality
13
Case Study: Automotive
www.mayato.com
Goals
• Optimization of production processes• Higher production quality• Efficient production planning across various production sites
Approach
• Structuring of data for import into the analytical data model• Recording of all production data in one central database• Integration of suitable parameters and limits for target-performance comparison
Results
• For the first time a complete and detailed view on all production data• Foundation for a unified and company wide reporting system• Higher quality through improved production monitoring• Near real-time allows for timely reaction
Quick Facts
14
www.mayato.com
• Automotive supplier
Industry
• SAS Business Analytics Platform• SAS Quality Lifecycle Analysis
Technology
• More than 10 million data points per system and day are recorded, stored and processed• Analysis of 350 million data records per month in one system alone• Integration of 12 shop-floor and final finishing systems• Planned rollout to 20 sites with 2000 users• Data volume in 2017 will reach 25 TB
Data volumes and performance
Quick Facts
15
www.mayato.com
Fragmented and disparate data silos
Valuable production information derived from shop-floor systems could not be used
No holistic view and analysis of production data
Delayed and solely reactive approach in case of production issues, quality problems or weak machine utilization
Initial situation
16
Results
www.mayato.com
Business perspective
• Continuously expand the usage of available data
• Expand maturity in regard to data usage
• From reactive data collection to optimized production
Technical perspective • Conception, design and implementation
of a central production management platform
• Integration of all systems and data assets that are used within the production process
• Central collection, visualization and analysis of production data in real-time
Solution
17
www.mayato.com
Business perspective
18
www.mayato.com
Technical architecture
19
www.mayato.com
Industry AnalyticsSave Energy, Buy Time: Automated Evaluation of Sensor Data Optimizes Steel Production
21
Case Study: Steel
www.mayato.com
Goals
• Optimization of resources consumption• Compliance with regulatory emission guidelines• Continuous quality controls
Approach• Integration, transformation and analysis of production, machine and
sensor data
Results
• Reduced operational costs• No payments of fines due to emission violations• Improved production processes and results
Quick Facts
22
www.mayato.com
Blast furnace data and additional sensor data are collected in various/different databases
Service personnel needs to check disparate data sources in order to get a valid production overview. This is associated with a high manual effort and waste of valuable time
Quality control is complex and inefficient
Operating costs are increasing in case of unchecked, defective sensor data (defect in temperature sensor can lead to an enormous waste of costly gas)
Initial situation
23
Results
www.mayato.com
Analysis of sensor and machine data
24
Emission
Model
Temp
Width
www.mayato.com
• Analysis of sensor data allows for early and deep insights into the production process
• Central collection, analysis and visualization of sensor data leads to:• Optimization of resources consumption• Compliance with regulatory emission guidelines and rules• Continuous quality controls
Results: Reduced operating costs, avoidance of costly fines, improved production processes
Summary
25
Eric J. EckerHead of Industry Analytics
Looking forward to your call:+49 160 98288828
… or your message:[email protected]
www.mayato.com
mayato GmbH • Headquarter: Am Borsigturm 9 • 13507 Berlin • Offices: Bielefeld, Mannheim, Vienna
Contact
26