The use of Physics Based Predictive Analytics in Digital Twins · •Physics based analytics offers...
Transcript of The use of Physics Based Predictive Analytics in Digital Twins · •Physics based analytics offers...
The use of Physics Based Predictive Analytics in Digital Twins
Frode Halvorsen, PhD
Vice President Technology
www.edrmedeso.com
Outline
• Business motivation
• Enabling technology
• Customer cases
IoT Value Chain
CustomerApplicationsPlatformConnectivitySensorsSmart
Product
+ +
Industrial IoT by EDRMedeso
Understand the health of asset
Analyze Real Time Data
Collect Data
Connect Asset
Tim
e to
dev
elo
p
Market Value
Condition Based Analytics Condition & Prediction Based Analytics
Use Historical Data to Predict Asset Behavior in the future
Understand the health of asset
Analyze the Real Time Data
Collect the Data
Connect the Asset
Market Value
Tim
e to
dev
elo
p
Analytics Solutions
Copyright EDR & Medeso AS 2016
Use-Cases Data DefinitionAlgorithm Definitions
ImplementationFill the Data
LakeModelling Product
3-10
Years
Analytics Solutions
Copyright EDR & Medeso AS 2016
Use-Cases Data DefinitionAlgorithm Definitions
Generate Data via Simulation
Algorithmic Modelling
Implementation Product
Weeks
Accelerated Physics Based Solutions
Unpredictive
Physical unit
Predictive
Digital twin
Realtime Future
Analysis
Sensor on physical unit
Trigger ROM
Full 3D
Predicting maintenance
and failure
Sensor on physical unit
Real time analyticsCondition based
“Traffic light indicators”
The Digital Twin
Copyright EDR & Medeso AS 2016
Ethernet
SmartTags Local Gateway(s) Cloud connect API
WiFi
Mobile
Mesh
BLE
Implementation
Copyright EDR & Medeso AS 2016
0
50
100
150
200
250
300
350
400
450
0 500 1000 1500 2000 2500 3000 3500
Vibration rms [mg]
Copyright EDR & Medeso AS 2016
Operational Output
WindVelocity
WindAngle
Rotational speed
Mechanical loads
Currentsweep
Displacements Damage?
Power
Positionsweep
Q
U
A
L
I
T
YVoltage/Current
M
A
I
N
T
E
N
A
N
C
E
Digital Prototypes
PMSM
free
a
b
c
mech_rv
Reduced-Order
Modeling
A simplification of a high-fidelity dynamical model that
preserves essential behavior and dominant effects,
for the purpose of reducing solution time or storage
capacity required for the more complex model.
How to get
from here…
…to here?
High-Fidelity 3D Physics
System-Level Behavior
Equation / Data-Based
Traversing Physical Model Fidelity
• 3 SVD modes only need to reconstruct the unsteady temperature field
• A temperature signal excitation is provided to cover the full temperature spectrum
• Verification of the Dynarom model created
310
320
330
340
350
360
370
0 2 4 6 8 10
Fluent
Dynarom
Time (s)
Seco
nd
ary
Inle
tTe
mp
erat
ure
(K)
Tem
per
atu
reFi
eld
(°C
)
ANSYS DYNAROM – Step 1
310
320
330
340
350
360
370
0 2 4 6 8 10
Seco
nd
ary
Inle
tTe
mp
erat
ure
(K)
Time (s)
Fluent
DynaromTe
mp
erat
ure
Fiel
d (
°C)
Step 2 – Arbitrary temperature field
ANSYS Twin Builder:
Controling TSR –Tip Speed Ratio
(Ratio between blade tip speed and wind speed)
Wind direction
Pitch angle – Control Signal
Rotor speed – Control feedback
Wind speed
Setpoint TSR
Digital Prototype
Digital Twins
ANSYS Digital Twins in the Cloud
Grundfos SYSMON digital twins - harness the power of IoT to optimize your life.Grundfos, a global leader in design and manufacturing of pumps and water systems, brings advanced simulation and prediction capabilities to real-time in collaboration with simulation-powerhouse ANSYS and elite channel partner EDRMedeso. Grundfos will use digital twins to better serve its customers through improved product quality and performance, enhanced development productivity, optimized maintenance and reduced overall costs and risks associated with unplanned downtime.
Main Takeaways
• The key technologies are ready; implementation can start tomorrow
• Commercial solutions available on the market
• Physics based analytics offers an accelerated process compared to most predictive analytics
• Established and proven methods can be applied in new areas
• Numerous use cases already implemented; first runners are running