[go: up one dir, main page]
Scribd
Upload
90 views5 pages

Github Implementationabhishek

This report details a machine learning framework developed to predict the mechanical behavior of composite materials using stress-strain data. The project, implemented in Python, combines finite element simulation data with machine learning to generate accurate stress-strain curves for various composite configurations. Results demonstrate that the trained models can effectively predict material behavior across different fiber volume fractions, offering a computationally efficient alternative to traditional simulations.

Uploaded by

ABHISHEk
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
90 views5 pages

Github Implementationabhishek

This report details a machine learning framework developed to predict the mechanical behavior of composite materials using stress-strain data. The project, implemented in Python, combines finite element simulation data with machine learning to generate accurate stress-strain curves for various composite configurations. Results demonstrate that the trained models can effectively predict material behavior across different fiber volume fractions, offering a computationally efficient alternative to traditional simulations.

Uploaded by

ABHISHEk
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Implementation Report: IGAPack-PhaseField

Name: Abhishek (Roll No: D24241)


Program: Ph.D. (IIT Mandi)

Abstract
This study presents a machine learning-based framework for accurately predicting the mechanical
behavior of composite materials, using stress-strain data and material properties.

Introduction
Purpose

To develop and validate a predictive model capable of constructing stress-strain curves for unseen
composite configurations with high accuracy.

About the Repository

The project is implemented in Python and focuses on developing a data-driven framework for predicting
the mechanical behavior of composite materials. The GitHub repository btp by ranamihir combines finite
element simulation data with machine learning techniques to generate stress-strain curves for various
composite configurations. It includes code, datasets, and visualizations to support accurate and efficient
prediction of material properties.

Repository Link: btp by ranamihir

Results for Stress-Strain Curve

The model that was trained on the stress-strain curve data was used to predict values (stresses) on
unknown composite materials for every volume fraction. This, in turn, was used to construct the
corresponding stress-strain curve.

A comparison has been made between the actual and predicted stress-strain curves for a particular
material at different volume fractions. The resulting graphs are as follows:
Figure 1: Actual and Predicted Stress-Strain Curves at 5% Volume Fraction

Figure 2: Actual and Predicted Stress-Strain Curves at 10% Volume Fraction


Figure 3: Actual and Predicted Stress-Strain Curves at 15% Volume Fraction

Figure 4: Actual and Predicted Stress-Strain Curves at 20% Volume Fraction


Figure 5: Actual and Predicted Stress-Strain Curves at 25% Volume Fraction

Figure 6: Actual and Predicted Stress-Strain Curves at 30% Volume Fraction


Conclusion

ML models trained on FEM data can accurately predict the behavior of composite materials across
varying fiber volume fractions.

Ensemble models (e.g., XGBoost + Random Forest) deliver better generalization and reduce overfitting.

The approach provides a computationally efficient alternative to full FEM simulations for new material
design.

References

Ciampaglia, A. "Data driven statistical method for the multiscale characterization and modelling of
fiber reinforced composites." Composite Structures 320 (2023): 117215.

You might also like