2. Phenomenological models
Phenomenological models simulate the observed behaviour without caring of the underlying physics.
They can represent relationships as stress-strain of fibres, force-displacement of rods, moment-rotation
of beam-to-column joints, base shear-top displacement of buildings, etc. without conceptual differences.
2AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
Some parameters may
not have clear physical
meaning
Model calibration
3. Model calibration
Model calibration means finding the model parameters which, inserted in the
numerical model of the a given test, provide the optimal fitting of the
experimental response.
3
S1
S2
S3
S4
Experimental
evidence
Real test
Numerical
modelling
Model calibration
AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
4. Multiple responses
4AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
We can define a discrepancy function, measuring the inconsistency between experimental data and
computed response:
𝜔𝜔 𝒑𝒑 =
𝒚𝒚𝒆𝒆𝒆𝒆𝒆𝒆 − 𝒚𝒚𝒄𝒄 𝒑𝒑
𝒚𝒚𝒆𝒆𝒆𝒆𝒆𝒆
Model calibration is thus solved as an optimisation problem
�𝒑𝒑 = arg min
𝒑𝒑∈𝑷𝑷
𝜔𝜔 𝒑𝒑
or, if NT responses are available,
�𝒑𝒑 = arg min
𝒑𝒑∈𝑷𝑷
𝜔𝜔1 𝒑𝒑 , … , 𝜔𝜔𝑁𝑁𝑇𝑇
𝒑𝒑
MULTI-OBJECTIVE OPTIMISATION PROBLEM
The general solution is the
Pareto Front,
the set of non-dominated solutions
Genetic Algorithms
5. MultiCal software - input
MultiCal
Experimental
tests
• Responses to fit
Hysteretic
model
• Parameters to
calibrate
Optimisation
settings
• Genetic
Algorithms
parameters
5AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
MultiCal (www.multical.unisa.it) addresses the need of a unique tool for
calibrating hysteretic models onto multiple experimental tests.
6. MultiCal software - input
MultiCal
Experimental
tests
• Responses to fit
Hysteretic
model
• Parameters to
calibrate
Optimisation
settings
• Genetic
Algorithms
parameters
6AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
MultiCal (www.multical.unisa.it) addresses the need of a unique tool for
calibrating hysteretic models onto multiple experimental tests.
7. MultiCal software - input
MultiCal
Experimental
tests
• Responses to fit
Hysteretic
model
• Parameters to
calibrate
Optimisation
settings
• Genetic
Algorithms
parameters
7AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
MultiCal (www.multical.unisa.it) addresses the need of a unique tool for
calibrating hysteretic models onto multiple experimental tests.
8. MultiCal software - output
MultiCal
Pareto Front
analysis
Compromise
solution
Curves
8AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
9. MultiCal software - output
MultiCal
Pareto Front
analysis
Compromise
solution
Curves
9AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
10. MultiCal software - output
MultiCal
Pareto Front
analysis
Compromise
solution
Curves
10AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
Moment-rotation Envelope Energy
11. Examples
11AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES
Repeated runs of the optimisation
analysis may increase the confidence in
the results and help assess uncertainty in
the calibration
14. Conclusions
• MultiCal is a tool aimed at calibrating hysteretic models based on multiple tests, in the framework
of multi-objective optimisation
• It can dealt with up to 6 tests, and different response quantities (generalised force, envelope curve
and energy)
• The optimisation is governed by Genetic Algorithms (NSGA-II)
• Several Opensees uniaxialMaterial models can be calibrated: more are planned to be implemented
• Default parameters and template builder assist the user in the creation of input files
• A graphical interface is under construction
14AUTOMATIC CALIBRATION OF HYSTERETIC MODELS THROUGH MULTIPLE RESPONSES