4. Coupling real and virtual assets Flashcards
CouplingMethodologies: Identification&estimation of states¶meters
How can we ensure that our virtual asset remains consistent with the physical asset? and how to find these parameters/inputs for the virtual asset?
For this we need the Inverse Problems and data assimilation: we find the parameters and the **states **from the real assets.
Parameters
* Some parameters are just known (e.g. from the geometric design)
‘Constitutive’ vs. ‘Geometric’ parameters
States
* Dynamics states (e.g. deformations in a component)
* Condition states
CouplingMethodologies: Identification&estimation of states¶meters
Stages of the inverse problem and data assimilation
CouplingMethodologies: Identification&estimation of states¶meters
2nd Step. Minimize the difference between the real and virtual asset, what do whe want to achieve
Minimize difference or minimize uncertainty? . This is done by generalized LEAST SQUARES PROBLEMS, two kind of solutions:
CouplingMethodologies: Identification&estimation of states¶meters
How to get the PARAMETERS/STATE/ESTIMATION of the system
Different aspects influence how well this can be done
* System (model) – the model is less important than the physics of the
system!
* Excitation
* Sensors
CouplingMethodologies: Identification&estimation of states¶meters
Which problems can we face with the IDENTIFICATION PROBLEM
We have different approaches:
*Time-domain
* Frequency-domain
* ‘feature’-domain
Main challenge: How to practically couple various types of models?
- Challenging to generally exploit models defined in commercial software in these methodologies
——FMI/FMU has the potential to streamline this! - Most of these algorithms require some hands-on programming still…
Estimators and Observers
What is the purpose of Estimators and observers
Obtain information from meassurements (sensors), fusion sensors
What is a linearobserver/estimator
Explain. Extended Kalman Filter!!
Difference between kinematic and dynamic model
Conclusions, Observers / Estimators
PARAMETER IDENTIFICATION
What is the difference between system identification VS parameter identification.
How to get the parameters?
What are the sources of information
System identification = To obtain the model of a system
Parameter identification = To obtain parameter values in that model
Offline parameters what represent and what online parameters.
What is the goal of the parameters in the virtual replica
Offline: represents the behaviour and online : the changes of the physical system
We want that the parameters of the model be as match as possible the parameters of the real systems
PARAMETER IDENTIFICATION
What is offline parameter identification, characteristics?
how to represent realistic behaiviour of the physical system: With good models + the right parameters values
Diffferent ways to run offline tests
Sine sweeping: apply a senosoidal signal and you meassure the respons (freq) dom. this is for the dynamics of the system.
PARAMETER IDENTIFICATION
What is the ON-LINE parameter identification, what do we want to reflect?—how to identify parameters of a system that is already in operation?
Here we want to REFLECT CHANGES in the model from the PS. if something happens in the PS here we should see them
Examples of identification in a Two mass systems, and how to identify the parametrs of G
How to represent a model, which function, how to find the parameters of the model when the system is online?
look for analogies, try to simulate with a two mass systems.
Look at the TRANSFER FUNCTION, this is the model!! this is the virtual replica. –Look at the parameters..Excite the system with PSEUDO RANDOM BINARY SIGNAL
PARAMETER IDENTIFICATION
When and why to use OFFLINE or ONLINE parameter identification
Digital Twin Verification and Validation
How to we check if there is consistancy between PA and VA?
-Focus on simulation models as Digital Twins
* The DT can consist of a number of models and configurations interacting
-Construct various models which can be used in different scenario’s to assess the behavior of the real asset
Digital Twin Verification and Validation
What is Verification and Validadation. What should be verificate and validate
Verification: does the DT meet the requirements, eg. is the model correctly implemented.
Validation::does the DT meet the needs of the user, eg. Can we extract the desired quantities of interest, is the model accurate enough to extract value
Digital Twin Verification and Validation
At which level and what should we verificate and validate
Model Element components and DT. Validate the algorithm and do test in different scenarios!.
Digital Twin Verification and Validation
Which set of test give us Strong validation and Weak validation?
Be aware to separate the data sets: Trainig/Validation/data set
Basics of uncertainty in Digital Twins
What is uncertainty in Digital Twins, where can there exists and which types