The document discusses digital twins, which are dynamic digital representations of physical assets that allow companies to understand, predict, and optimize asset performance. Digital twins use asset data like sensor readings, events, and models to generate insights about an asset's current context, key performance indicators (KPIs), and future predictions. A digital twin platform is needed to manage digital twins at scale across edge, network and cloud environments and expose twin data and insights via APIs. This allows industrial applications to leverage digital twins without needing direct access to the underlying data and models.
1. The Digital Twin:
A Radical New Approach to IoT
IOT DevCon 2017
Dimitri Volkmann, Digital Twin Thought Leader, GE Digital
www.linkedin.com/in/dimitrivolkmann
@dimiexter
www.pinterest.com/dimitrivolkmann/digital-twin/
2. Economic Power is Shifting
(Credits: Geoffrey A. Moore – published April 12, 2017 on LinkedIn & Twitter)
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
4. Digital twins are dynamic digital
representations that enable
companies to understand,
predict and optimize the
performance
of their machines
and their business.
Digital twins are a persistent
digital model of the structure,
behavior and context of a
physical industrial asset.
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
5. The Digital Twin Canvas
Digital twins are dynamic digital representations that enable companies to understand, predict
and optimize the performance of their machines and their business.
Asset Data , Metadata & Events Insights (Industrial Intelligence)
Structure
• CAD Model & Specs
• BOM
• Asset Model
• …
Behavior
• Sensors reading
• Events
• Control Systems
Gold Data
Training Data
Models &
Orchestration
Context
• Environment
• Configuration
• History (Service, etc)
• …
Analytics Kernels(physics)
Specific Outcomes
Early warnings
Predictions
Optimizations
Machine Learning
(Model Training)
In Out
DataFederation
Data&MetadataManagement
Predix Platform
Asset
Connectivity
Data
Persistence
Model
Execution
Continuous
Learning
Machine Learning
(Auto Data Modeling)
Edge to Cloud
data & compute
Digital Twin APIs (Context, KPIs, Insights)
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
6. Digital Twin: Technology Value
Context Data – The Past
3D model, eBOM, Service
Records, First Date of
operation…
KPIs – The Present
Sensors Reading,
Performance, Current
State…
Insights – The Future
Failure Modes, Early
Warnings, Predictions…
www.formula1.com
Context
KPIs
Insights
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
7. Asset-centric, Industrial Application Types
- Monitoring &
Health
- Diagnostics
- Maintenance
Optimization
- Reliability
Management
- Performance
Optimization
- Compliance
- Operations
Optimization
- Asset Strategy
- Business
Optimization
- New Business
Models &
Services
Physical
System
Digital
Twin
Individual
Asset level
Operations/
Enterprise
level
- Simulations
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
8. Digital Twin Development Paradigm
BUILD; MANAGE, RUN, PERSIST & LEARN; CONSUMME
1
2
3
BUILD: Asset Experts team up with data management, data scientist, analytics and
machine learning specialist to BUILD Digital Twin Classes – for a specific class of Asset
or System – in order to assemble data and models for identified desired outcomes.
These Digital Twin Classes are published in a catalog.
RUN: Digital Twins instances for identified Assets or Systems are persisted, executed
and optimized on the Platform. As more Digital Twins run on the Platform, the
Learning System improves the Digital Twin Models. The Platform also offers Digital
Twin management.
CONSUME: Digital Twin offers Context Data, KPIs and Insights about an Asset or
System. This value is accessible through APIs and is used to build Industrial Internet
Apps such as APM from GE Digital.
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
9. Deployment: Edge to Cloud
CLOUD EDGE NETWORK
• Cheapest
• Very Big Data
• Very Big Compute &
Machine Learning
• Compliance &
Auditability
• Lowest Latency
• Lots of low relevance
data (filtering)
• Resilience
• Governance
• Many to many
interactions
Most deployment will use a distributed model Edge/Network/Cloud
depending on specific requirements, the Platform can self adjust based on learning
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
10. Why a Digital Twin Platform?
• “Where Digital Twins live”
• Data and IP Governance
• Monetization of Data, IP and
Outcomes of the Digital Twins
• Learning System
• Scale
• Management & Compliance
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
11. The RDBMS Analogy
RDBMS DIGITAL TWIN PLATFORM
Digitalize Information Systems & records Digitalize Physical System
Tables, Index, Clusters, Stored Procedures Digital Twin Class
Records Digital Twin Instances
SQL: DDL (Tables & Views) and Stored
Procedures
Digital Twin Class (Data, Metadata, Analytics, Models,
etc)
SQL: SELECT, INSERT, UPDATE, DELETE APIs Calls for Context, KPIs, Insight
RDBMS Engine: persist data, optimize queries Digital Twin engine: persist data, optimize Analytics,
Learning System
Database Designer, Database Analyst Asset Experts, Digital Twin Class Designers
Database Developers Industrial Apps Developers (use APIs)
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
12. Getting Started Today with Digital Twins: www.predix.io
Design Principles
• Manage the Data, expose
relevant Data (past,
present)
• Build Models & Analytics for
re-use
• Orchestrate to deliver
specific outcome: KPIs, early
warning, predictions
• Expose as API (de-couple
from UI/UX)
GE Digital 2017, no Reproduction – IoT DevCon Conference April 2017
13. IOT DevCon 2017
Dimitri Volkmann, Digital Twin Thought Leader, GE Digital
www.linkedin.com/in/dimitrivolkmann
@dimiexter
www.pinterest.com/dimitrivolkmann/digital-twin/
Thank You!