Machine learning-based prediction of mechanical properties of viscoelastic dampers

This study explores the application of machine learning (ML) models, including Random Forest (RF), Extreme Gradient Boosting (XGB), and Artificial Neural Network (ANN), for predicting the mechanical properties of viscoelastic dampers (VED), specifically the storage and loss modulus. VEDs play a crucial role in structural engineering for mitigating dynamic responses to seismic and wind forces. Despite their effectiveness, predicting the mechanical properties of VEDs remains a challenge due to their sensitivity to various factors such as loading amplitude, frequency, and temperature. Leveraging ML models and Explainable AI (XAI) techniques, this research aims to enhance understanding of VED behavior under cyclic loading and provide valuable insights for utilizing ML models for prediction of VED mechanical properties. The study conducts experiments within a temperature chamber, subjecting VEDs to cyclic loading in different conditions to discern the effects of these features on storage and loss modulus. The features such as loading amplitude, frequency, temperature, and loading cycle are then utilized to train ML models. XAI techniques are applied to provide insights into the predictive mechanisms of these models, ensuring the accuracy and reliability of predictions. The results indicate that all three ML models exhibit commendable prediction capabilities, with the ANN demonstrating superior performance compared to RF and XGB. According to SHAP (Shapley Additive Explanations) analysis reveals that loading amplitude exhibits the highest impact, followed by working temperature, loading cycle, and loading frequency.