Machine Learning in Fiber Composites

his lecture package provides a comprehensive exploration of an innovative approach to predicting the carbon fiber properties using a machine learning-based inverse model. The course is structured into several key sections:

1. Introduction to Carbon Fiber Properties:
– Overview of the challenges in experimentally determining the carbon fiber properties.
– Explanation of how these properties are traditionally estimated using micromechanical models.

2. Machine Learning Framework:
– Introduction to Gaussian Process Regression (GPR) and its application in creating a surrogate model.
– Discussion on how GPR not only predicts fiber properties but also quantifies uncertainty in the predictions.

3. Data Generation and Model Training:
– Explanation of how synthetic data is generated using Finite Element (FE) homogenization, considering various fiber and matrix properties, volume fractions, and fiber distribution.
– Steps involved in training the GPR model using this synthetic data.

4. Inverse Modeling Approach:
– Detailed walkthrough of the inverse approach used to predict the elastic properties of polyacrylonitrile (PAN)-based fibers in carbon-epoxy composites.
– Comparison of the model’s predictions with experimental values from literature and demonstrating its accuracy .

5. Validation and Robustness:
– Discussion on the model’s validation against experimental data for composites not included in the training set, achieving a regression coefficient above 0.93.
– Examination of the model’s robustness in the presence of uncertainty and noise.

6. Practical Application and Case Studies: Multi-Scale Homogenization and Woven Fabrics:
– Practical exercises and case studies demonstrating how to apply the ML-based inverse modeling approach in real-world scenarios.
– Case studies on different weave patterns (Plain, Twill, and 5 Harness Satin) to illustrate the model’s versatility.

7. Conclusion and Future Directions:
– Summary of the benefits of using a GPR-based surrogate model for predicting fiber properties.
– Exploration of potential future applications and improvements in the field of composite material analysis.

This package is designed for advanced students and professionals in materials science and mechanical engineering, providing them with both theoretical knowledge and practical skills in applying machine learning to composite material analysis.