Machine Learning Applications in Civil Engineering

This research topic explores the use of soft computing methods, such as genetic algorithms and artificial intelligence, to develop modern pavement indices for road networks in Jordan, as well as the bond behavior between corroded/cracked reinforced concrete and NSM strips, and the potential of blockchain technology to revolutionize the sustainability of energy transitions. Additionally, machine learning techniques are used to analyze the risk factors that affect crash severity levels, and a machine learning framework is proposed for calibrating the parameters of analytical models of complex nonlinear structural systems.

The results of the pavement indices study showed an efficient performance benefit of using these techniques, with the ANN and GEP models able to predict the output variable with a reasonable accuracy. The bond strength between corroded/cracked reinforced concrete and NSM strips was found to be significantly decreased with increasing corrosion level, with a maximum reduction of 48%. Blockchain-enabled P2P trading was found to have lower processing trading transaction costs than current coordination costs. Association rule, random forest, decision tree, and AdaBoost algorithms were applied to predict crash severity levels and investigate the impact of drivers, vehicles, and road characteristics on traffic crashes.

The proposed machine learning framework for calibrating the parameters of analytical models of complex nonlinear structural systems had a convolutional neural network architecture. Additionally, the performance of different forms of square steel plates with a b/h ratio of 100 was investigated, and it was found that stiffened plates had a significantly higher maximum strength than intact plates. Finally, the GeneXproTools program was used to analyze a dataset related to slope stability circular failure cases, with the proposed models having an accuracy of 93.1% and an R2 value of 0.96 when compared to the available models in the literature.