Predictive Models to Evaluate the Interaction Effect of Soil-Tunnel Interaction Parameters on Surface and Subsurface Settlement

Nowadays, the need for subway tunnels has increased considerably with urbanization and population growth in order to facilitate movements. In urban areas, are excavated shallow depths under densely populated areas soft ground. Its associated hazards include poor ground conditions surface settlement induced by tunneling. Various sophisticated variables influence of caused The shield machine's operational parameters critical due complexity shield-soil interactions, tunnel geometry, local geological parameters. Since all elements appear have some effect on tunneling-induced settlement, none stand out as particularly significant; it might be challenging identify most important ones. This paper presents a new model an artificial neural network (ANN) based partial dependency approach (PDA) optimize lack explainability ANN models evaluate sensitivity response tunneling prediction subsurface settlement. For this purpose, 239 104 points monitoring respectively, were obtained from line Y, west bond Crossrail London. surface, trough, boring machine (TBM) used categorize 12 potential input that could impact maximum An standard statistical multiple linear regression (MLR) also show capabilities PDA displaying parameter's interaction impact. Performance indicators such correlation coefficient (R2), root mean square error (RMSE), t-test generated measure performance described models. According results, geotechnical engineers general practice should attend closely index properties reduce risks related results revealed two different effects target parameter change overall entire model. Remarkably, between was observed more precisely zone than zone. comparison indicated proposed PDA-ANN is reliable MLR presenting It can further applied establish multivariate consider single model, better capturing among parameters, leading realistic demand assessments. study will benefit underground excavation projects; experts make recommendations criteria control controlling predicted results. Doi: 10.28991/CEJ-2022-08-11-05 Full Text: PDF