Pop-In Identification in Nanoindentation Curves with Deep Learning Algorithms

Area of Science:

• Materials Science
• Data Science
• Artificial Intelligence

Background:

• High-speed nanoindentation generates large datasets requiring advanced analysis.
• Differentiating complex load-displacement curves (e.g., with pop-ins) from typical ones is challenging.

Purpose of the Study:

• To develop and validate an AI model for classifying nanoindentation load-displacement curves.
• To automate the identification of specific curve features like pop-ins.

Main Methods:

• A convolutional neural network (CNN) model was implemented using Python, TensorFlow, and Keras.
• Load-displacement curves from various materials were converted into 50x50 matrices for model input.
• The CNN model was trained and validated on datasets with and without pop-in events.

Main Results:

• The CNN model achieved approximately 93% accuracy in differentiating pop-in from non-pop-in curves.
• The model demonstrated a negligible risk of overfitting.
• Successful classification across diverse material datasets was achieved.

Conclusions:

• AI, particularly CNNs, offers a powerful tool for analyzing large nanoindentation datasets.
• Automated classification of load-displacement curves enhances materials characterization.
• Computer vision models can effectively interpret complex nanoindentation data patterns.