Automated Prediction of Crack Propagation Using H2O AutoML

Area of Science:

• Materials Science and Engineering
• Computational Materials Science

Background:

• Crack propagation significantly impacts structural integrity and component durability.
• Accurate prediction of crack behavior is crucial for engineering reliability and safety.
• Existing methods for crack prediction demand more efficient and automated solutions.

Purpose of the Study:

• To develop and evaluate a machine learning-based automated model for predicting crack propagation.
• To leverage the H2O library for analyzing crack patterns and forecasting behavior.
• To enhance structural health monitoring and engineering safety through accurate crack prediction.

Main Methods:

• Developed an automated model using the H2O library.
• Utilized a dataset of crack propagation instances in Acrylonitrile Butadiene Styrene (ABS) specimens.
• Employed rigorous evaluation metrics (MAE, RMSE, R²) and cross-validation for accuracy and robustness assessment.

Main Results:

• The automated model demonstrated remarkable accuracy and reliability in predicting crack propagation.
• The H2O library proved effective for analyzing complex crack patterns.
• Cross-validation confirmed the model's generalizability across diverse datasets.

Conclusions:

• Automated Machine Learning (AutoML) offers a powerful approach for predicting crack propagation.
• The H2O library is a valuable tool for structural health monitoring applications.
• This research advocates for the adoption of AutoML in engineering for improved safety and reliability.