predictive quality management - sap...
TRANSCRIPT
Use this title slide only with an image
Predictive Quality Management Optimize manufacturing operations with SAP Predictive Analytics and SAP HANA
Public
© 2014 SAP AG or an SAP affiliate company. All rights reserved. 2 Internal
Predictive Analytics: Typical use cases
q Prospect
q Cross-sell & Up-sell
q Churn
Marketing
q Credit risk
q Insurance risk
q Fraud
Finance
q Predictive Maintenance
q Warranty Analysis
q Demand forecast
Industry
© 2014 SAP AG or an SAP affiliate company. All rights reserved. 3 Internal
q Customers q Products q Transactions q Equipments
q Appetency q Risk q Fraud
q Marketing q Finance q Maintenance
Historical data Typical and recurrent behaviors Predict and act
Predictive Analytics concepts
LEVERAGE DATA FOR CONTINUOUS PROCESS IMPROVEMENT
Get the Data Securely connecting potentially millions of devices/sensors to a
central place, collecting enormous amounts of data
Automate Actions Provide platform to create and run
value-added applications, integrated with all IT assets
Insight from Data Enable real-time analysis of vast
amount of data applying consistent predictive algorithms
leveraging sensor and other structured and unstructured data
SENSE All relevant variables (temperature, humidity, pressure, etc.)
2_1 2_2
2_3
INTEGRATE Integrate captured data (like image recognition) plus IT data
ANALYZE Automated analysis of error patterns PREDICT
Real-time prediction of quality results
IMPROVE Incorporate insights into production process
COLLECTING AND MONITORING SENSOR DATA
Streaming data collected every second from sensors
SAP HANA real-time data
platform
SAP BI real-time
dashboards
Position
Vibration
Temperature
Pressure
Humidity
10101010101 01000101001 10010110110
1. Sensors measure activity indicators every second
2. Sensor data is collected and stored in real-time into the SAP HANA platform
3. Users can monitor sensor activity in real-time using SAP BI Dashboard
4. Big Data produced by sensors can be archived into low-price Hadoop
BUILDING MODELS FROM SENSOR DATA
SAP HANA real-time data platform
SAP Hana Predictive Libraries
SAP Predictive Analytics
1. Data scientists transform IoT raw data into analytical data with SAP Predictive Analytics Data Manager
2. They can easily access to Hadoop historical data and merge it with HANA data using SAP HANA Smart Data Access
3. SAP Predictive Analytics Modeler leverages SAP HANA Predictive libraries and SAP Spark Predictive libraries to avoid data duplication and enable in-memory model training
Data scientist SAP Spark Predictive Libraries
REAL TIME PREDICTIONS WITH SAP HANA
Streaming data collected every second from sensors
Position
Vibration
Temperature
Pressure
Humidity
10101010101 01000101001
1. SAP Predictive models are deployed into the HANA streaming process
2. They will predict events such as potential upcoming failures in real-time
3. Alerts can be raised and sent to business users through various channels
Predictive Models
SAP Mobility
SAP HANA real-time data
platform
SAP BI real-time
dashboards
SAP Digital ManufacturingValue Creation through Digitization
Eric ThierenEMEA Solution Sales FBS, Manufacturing Industries
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 2Internal
Trends Impacting Digital ManufacturingDigital transformation in the extended supply chain
Sharing EconomyBusiness Network
Resource ScarcitySustainability
DesignRe
spon
dEnd-to-EndVisibility
Customers &Markets
Suppliers &Logistics
Products &Assets
People &Resources
Individualized ProductsIndustry 4.0
Customer CentricityOmni-Channel
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 3Internal
Connecting Things with People and ProcessesSAP Leonardo Innovation Portfolio
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 4Internal
SAP Digital Manufacturing is ConnectedEnd-to-End Business Process and Business Partner Integration
Seamlessly Integrated Business Processes: Top 5 Scenarios
• Top Floor to Shop Floor
Live response in production planning and scheduling for high flexibility with lastminute demand changes
Manufacturing execution & orchestration of the automation layer in lot size of 1
• Machine-to-Machine communication
• E-Commerce integration: Sales order entry with configuration andpersonalization for individualized products
• Machine Cloud integration
• Direct replenishment, Outsourcing
Insight-to-Action based on machine & business data for high transparency
ServiceProvider
E-CommerceConsumer/Customer
MachineCloud
ERP
SAP ManufacturingExecution Suite
Supplier
Top Floor
Shop Floor
1
2
3
4
5
1
2
3
5
4
34BusinessNetworks
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 5Internal
Digital Manufacturing InsightsInsights to Action @ All Levels
CxO
Supervisors
Review Revenue, Cost, Delivery Performance, Quality &Customer satisfaction across manufacturing plants and takedecisions
KPI Score Card
Monitor and Analyze Manufacturing Performance
real time production insights and expedited actions
Continuous improvement
Real time visibility of manufacturing
Drive improvement with insight to action toimprove core KPI’s
Flexibility to untether from the desk
Key PerformanceIndicators
ManufacturingPerformance
Indicators
ManufacturingActivity Indicators
PeopleQuality Cost Delivery
Planned Innovation
Operators
Devices and Machines
Safety
Guidance &Real-timefeedback
Automation &PredictiveAnalytics
Clear and consistent guidance
Real-time feedback
Reduced operator burden of non-value adding tasks
Automation of data
Better information
Device to Enterpriseconnectivity
Plant &Unit Heads
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 7Internal
Industry 4.0Digitization has big pay back
PerformanceCost -7%Net Margin + 19%Dividend +25%Share Repurchase >2.5MMDemand +40%From 21 day schedule to 6 hour
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 8Internal
Industry 4.0 in Paper ProductionReal Time Costing – Energy Management
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 9Internal
Digital Operations in Glass MakingReal time visibility across the value chain
Architectural glassMonitor raw material, through batching, float glass and finishingMaterials monitoring for qualityEnergy analysisEmbedded solar panels create IOT opportunity
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 10Internal
Packaging and Palletizing AutomationPallet Labeling Provides High Return on Investment
© 2016 SAP SE or an SAP affiliate company. All rights reserved.
Thank youContact information:
Eric ThierenFocused Business Solutions Sales, EMEA,Discrete [email protected]+32499513726
Philippe Nemery, SAP BeLux
Bi and Predictive Presales - Presales Manager Platform Solutions Group
Predictive Quality ManagementCustomer References.
4INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
5INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Koehler Paper Group
Koehler Paper Group – one of the very fewindependent and family owned papergroups in Europe
7 paper machines and 1 board machine at 4 locations
• Sales Volume 2014: 500.000 t
• Sales Value 2014: 650 Mio. €
• Employees 2014: 1.694
Papierfabrik August Koehler SE
Oberkirch
Koehler Kehl GmbH
The company is a producer of high-quality special papers.
6INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
7INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Predictive Models for Production Quality ParametersProject Overview
Goals
Reduce time-to-quality on product switchover
Monitor and guarantee quality standards continuously
instead of through one lab measurement at the end
Understand and verify central drivers for quality
parameters
Approach
Pilot phase: built predictive models for 6 fundamental
quality parameters using SAP PA
Usage of type-blending models to overcome small
training data sizes and adaptation to first-time
materials
Model export to SAP HANA DB for real-time
prediction of quality parameter
Challenges
Few training data due to relatively expensive lab tests
for ground truth
Large variety of materials with different quality goals
Data from waste production is not recorded (lack of
variance in training data)
Outcome
Model evaluation phase indicated sound reliability of
quality predictions for regularly produced products
Correlation analysis confirmed production expert’s
conjectures about quality drivers
Bias of one hardware sensor detected
Software training to enable continuing model
improvements by customer using newly available data
8INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Predictive Models for Production Quality ParametersModel Performance
Average Model fit (blue)
to validation data (green)
Typical model variance
(1 sigma area)
Model prediction (red)
vs. lab results (blue)
Model performance assessment
PA model built for each quality parameter based on training
set with ~500 samples
Model validation showed typical model variance well within
acceptable tolerances for all relevant parameters (top chart)
Implementation of predictive models in SAP HANA for real-
time quality prediction
Model evaluation phase for ~1 month showed sound match
with lab results (right chart)
9INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Predictive Quality: Paper Producer
Benefits
• Reduce waste and rework
• Faster “Time-to-Quality” for small lot sizes and first time materials
• Superior product quality
• Detailed automated quality documentation
Process Innovation
Data Selection
Visualization
Statistical Analysis
Expert Review Step-by-
step root
cause
analysis
“Predictive Quality means we calculate in real-time what the quality of the product that is currently in production will be.”
10INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
11INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Devices
BO Dashboards
External
DataFinancial,
Sales
Big Data Lake Training of Predictive Models
SDS
Triggering of alarms/actions
based on predictive models
SDS: Smart Data Streaming
SDI - SDQ
Hana Vora
Kafka &
Zookeeper
Real Time
Gateway
System
Real-time Stream
Analysis
Visual
Exploration
SAP HANA
Platform
HDFS SparkHIVE
16INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
© SAP 2009 / Page 16
Predictive Analytics InsightsFocus on a single Predictive Analytics Use Case that is
time-boxed for rapid delivery and directional insights (2) weeks
Design and pilot deployment:
(4-6) weeksData Science Solution Design and
pilot Deployment
Data Science Production
Iterations & Customer
Enablement
Deploy
Increment 1
Go or no-
go decision
Predictive analytics
(PA) Go or no-go
decision
Predictive Process: from test and pilot to deploymentData Science team
Deploy
Increment 2
Deploy
Increment 3
Deploy
Increment n
(8-12)
weeks
17INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Predictive Team
Strategic Manager
o Defines the business strategy.
o Subject-matter expert.
o Judges the value and the ROI.
(Citizen) Data Scientist
o Creates the data sets.
o Creates quality models.
o Keeps models up-to-date.
Business Requirements
Line of Business
Predictive Project Team
Business-Relevant Insights
IT Support
IT Operations
o Give (Access to) Data
o Installs products
o Manages security
Support
18INTERNAL© 2017 SAP SE or an SAP affiliate company. All rights reserved. ǀ
Predictive Team
Strategic Manager
o Defines the business strategy.
o Subject-matter expert.
o Judges the value and the ROI.
(Citizen) Data Scientist
o Creates the data sets.
o Creates quality models.
o Keeps models up-to-date.
Operational Decision Maker
o Consumes the information.
o Perform recommended actions.
Data Architect
o Federates the data sources.
o Builds data models.
o Integrates model-generated data.
Project Lead
o Drives the predictive project.
o Orchestrates the different roles.
o Make sure of the business value.
Developer
o Integrates the predictive project results in front-ends, applications, or business-intelligence reports.
Business Requirements
Line of Business
Predictive Project Team
Business-Relevant Insights
IT Support
IT Operations
o Operates servers
o Installs products
o Manages security
Database Admins
o Manages databases
o Defines access rights
Support