Automatised selection of load paths to construct reduced-order models in computational damage micromechanics: from...

Area of Science:

• Computational micromechanics
• Model order reduction
• Parameter space analysis

Background:

• High dimensionality of parameter spaces in micromechanics poses challenges.
• Selecting representative data (snapshots) is crucial for model accuracy.
• Existing methods struggle with exhaustive snapshot set selection.

Purpose of the Study:

• To present novel, reliable model order reduction (MOR) strategies.
• To address the challenge of selecting exhaustive snapshot sets in computational micromechanics.
• To enable the construction of accurate reduced-order models (ROMs).

Main Methods:

• Utilized random sampling of energy-dissipating load paths.
• Employed Bayesian optimization with interlocked parameter space division.
• Focused on generating an exhaustive snapshot set.

Main Results:

• Successfully developed reliable MOR strategies.
• Demonstrated effective selection of exhaustive snapshot sets.
• Validated the ability to build reliable reduced-order models from selected snapshots.

Conclusions:

• The proposed strategies ensure exhaustive snapshot set selection.
• Reliable reduced-order models can be constructed for computational micromechanics.
• Advanced sampling and optimization techniques overcome parameter space challenges.