Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Machine learning for development of constitutive model for iron-based shape memory alloys
Recent developments in the use of shape memory steel in structural engineering has highlighted the need for a better model. In parallel machine learning has ben very successfully used to identify constitutive laws, hyperelastic, plastic. This project proposes to apply the ML approach to Fe-SMA.
In recent years, a new iron-based shape memory alloy has been developed at Empa. It already has field applications and research on the Fe-SMA is still ongoing. The main application of this alloy is structural strengthening, indeed the prestressing possibilities offered by the SMA make it a very interesting candidate for the development of retrofitting elements. Such retrofit/repair solutions have been identified in the International Energy Agency report as a mean to reduce the energy use and carbon emission of the construction industry.
The need for a proper model of the Empa Fe-SMA has been identified in the recent months. Thermomechanical models for SMA exist in the literature, while most are focused on NiTi alloys, that exhibit limited plasticity, there is also one model (Khalil 2012) for Fe-SMA that includes phase transformation (the base for shape memory effect) and plasticity.
The project proposes to identify key parts of this current existing model in order to improve its accuracy and identify a constitutive model of the studied Fe-SMA.
The current model is developed in part using a decomposition of the elastic potential energy. 4 power laws are used with 8 material properties are used. This project aims at identifying not only the parameters but the the ML approach allows to look for the best function and not only its parameters. This approach has been successfully applied for hyperelastic and plastic constitutive models.
It is believed that such a model would be of significant value for the field of Fe-SMA and would help expand the use of this retrofitting technique.
For a more detailed description of the proposal and references please refer to the document attached.
The project would be supervised by an ETH or EPFL professor along with a member of Empa
In recent years, a new iron-based shape memory alloy has been developed at Empa. It already has field applications and research on the Fe-SMA is still ongoing. The main application of this alloy is structural strengthening, indeed the prestressing possibilities offered by the SMA make it a very interesting candidate for the development of retrofitting elements. Such retrofit/repair solutions have been identified in the International Energy Agency report as a mean to reduce the energy use and carbon emission of the construction industry.
The need for a proper model of the Empa Fe-SMA has been identified in the recent months. Thermomechanical models for SMA exist in the literature, while most are focused on NiTi alloys, that exhibit limited plasticity, there is also one model (Khalil 2012) for Fe-SMA that includes phase transformation (the base for shape memory effect) and plasticity.
The project proposes to identify key parts of this current existing model in order to improve its accuracy and identify a constitutive model of the studied Fe-SMA.
The current model is developed in part using a decomposition of the elastic potential energy. 4 power laws are used with 8 material properties are used. This project aims at identifying not only the parameters but the the ML approach allows to look for the best function and not only its parameters. This approach has been successfully applied for hyperelastic and plastic constitutive models.
It is believed that such a model would be of significant value for the field of Fe-SMA and would help expand the use of this retrofitting technique.
For a more detailed description of the proposal and references please refer to the document attached.
The project would be supervised by an ETH or EPFL professor along with a member of Empa
Part I : Unidirectional proof of concept
- From unidirectional tensile tests, identify fitting points for the ML model and propose a 1st estimate of the
- Fit the existing model (Khalil 2012) to the unidirectional data
- Propose a first set of identified functions based in the unidirectional data
Part II: Experiments
- Prepare and perform the 2D experiments
- From DIC displacement fields decouple each strain field along with the 2 scalar fields
Part III: Identification and validation
- Identity the 2D constitutive law with the ML fit
- Validate the model in FEM
Part I : Unidirectional proof of concept
- From unidirectional tensile tests, identify fitting points for the ML model and propose a 1st estimate of the
- Fit the existing model (Khalil 2012) to the unidirectional data
- Propose a first set of identified functions based in the unidirectional data
Part II: Experiments
- Prepare and perform the 2D experiments
- From DIC displacement fields decouple each strain field along with the 2 scalar fields
Part III: Identification and validation
- Identity the 2D constitutive law with the ML fit