Crystal Plasticity Parameter Optimization in Cyclically Deformed Electrodeposited Copper-A Machine Learning Approach.

Area of Science:

• Computational Materials Science
• Machine Learning Applications
• Materials Engineering

Background:

• Parameter optimization is crucial for accurate material modeling.
• Traditional methods can be time-consuming and require extensive experimental data.
• Elasto-viscoplastic models with isotropic and kinematic hardening present complex optimization challenges.

Purpose of the Study:

• To apply a machine learning approach for efficient parameter optimization in material models.
• To validate the method for elasto-viscoplastic models under cyclic deformation.
• To demonstrate the ability to obtain material parameters without further optimization runs.

Main Methods:

• Utilized a machine learning approach based on long short-term memory (LSTM) networks.
• Applied the method to an elasto-viscoplastic model incorporating isotropic and kinematic hardening.
• Evaluated performance for cyclic deformation in the low-cycle fatigue regime.

Main Results:

• Achieved reasonable agreement of stress-strain curves for cyclic deformation.
• Demonstrated the capability to obtain parameters for new materials without re-optimization.
• Validated the method's robustness on complex problems like crystal plasticity and self-consistent models.

Conclusions:

• The proposed LSTM-based machine learning method offers an efficient alternative for material parameter optimization.
• The approach significantly reduces the need for repeated optimization processes for new materials.
• The method's demonstrated power and robustness make it broadly applicable to various experiments and material models.