Predicting Steel Fatigue Using Machine Learning Techniques for sustainable Infrastructure Monitoring

Ensuring the structural integrity of steel infrastructure is crucial for sustainable development. Fatigue damage, a major concern in steel structures, can lead to catastrophic failures. This paper investigates the application of machine learning (ML) techniques for predicting steel fatigue life. The work compares the performance of eight regression models using a publicly available dataset MatNavi from the National Institute for Material Science (NIMS). The model performance is evaluated using three metrics: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared Score. The results show that Gradient Boosting achieves the lowest error metrics, demonstrating superior performance in predicting fatigue life. Random Forest also exhibits high accuracy, suggesting its potential as an alternative method. While Huber Regressor shows higher error values, its potential for handling outliers might be valuable in specific scenarios. This study highlights the effectiveness of machine learning, particularly ensemble methods, for predicting steel fatigue and emphasizes their potential in enhancing sustainable infrastructure management.