Implementation of Predictive Machine Learning Algorithms on Heated Geothermal Bridge Data

In recent years, there has been a notable surge in exploring statistical methodologies and artificial intelligence (AI) approaches, such as machine learning and deep learning, across various domains, including engineering. These data-driven techniques hold promise for delivering faster and more accurate predictions, representing a significant avenue for progress. Building upon prior research, this study extends the scope by incorporating additional features and utilizing a distinct dataset for forecasting the average heated bridge deck surface temperature and the outlet fluid temperature from the hydronic heating loops. Leveraging machine learning algorithms on data collected from a bridge de-icing project in Texas, we examine the effectiveness of Multiple Linear Regression (MLR) and Support Vector Regression (SVR). Through rigorous comparison with field data, we validate the robustness of these methodologies in temperature forecasting tasks, noting high accuracy across all algorithms. Notably, while both MLR and SVR demonstrate commendable performance, MLR marginally outperforms SVR, achieving an R
²
value of 0.79.