Artificial Intelligence to Predict Maximum Surface Settlements Induced by Mechanized Tunnelling

Among the construction methods developed for tunneling, mechanized excavation by Tunnel Boring Machines (TBMs) is currently considered a preferred option for technical and safety reasons in an urban environment, where damage induced on pre-existing building and services should be minimized. Since the ability to predict TBM performances is a critical point required to enhance the quality of the excavation and to optimize time, cost and safety operations in a project and since real-time prediction should be done during excavation in order to adjust some parameters in very real-time, approaches based on Artificial Intelligence (AI) methodology could be crucial. This study proposes an expeditious tool based on the application of Artificial Intelligence and particularly Artificial Neural Networks (ANNs), to predict the maximum surface settlements induced by tunnelling. ANNs, taking advantage of the quality of data available and computational performances of software for data management, have been proved to be a reliable instrument in processes where a relevant number of parameters and acquired measurements have to be managed. Using data selected from the excavation of the Milan M5 metro line, the document includes details on the role played by several inner elements on the accuracy of the final prediction based on the comparison of several different ANN configurations. The obtained results showed a promising capability of the tool to swiftly predict surface settlements in mechanized tunneling projects.