Designing tunnels in liquefiable sandy soils presents a significant challenge in determining the optimal depth and extent of soil cementation around them. Reducing the depth of the tunnel decreases both the bending anchor force and the axial load on the tunnel's shell,
More
Designing tunnels in liquefiable sandy soils presents a significant challenge in determining the optimal depth and extent of soil cementation around them. Reducing the depth of the tunnel decreases both the bending anchor force and the axial load on the tunnel's shell, yet it leads to an increase in ground surface settlement, and the opposite is true when depth is increased. Enhancing the cementation level at the tunnel's optimal depth reduces both structural uplift and shear forces exerted on the tunnel lining, but it also leads to an increase in axial loads and vice versa. Given the contradictory nature of these outcomes, the FLAC software was employed to simulate tunnels in liquefiable soils to address this dilemma. Subsequently, a neural network was utilized to identify correlations between the inputs and outputs of the simulation. This network was the objective function for identifying optimal values by applying a genetic algorithm. Optimal design parameters were derived using the NSGA-II modified algorithm, a multi-objective optimization technique based on the objective functions. Ultimately, Pareto charts generated from the multi-objective optimization process enabled designers to select the most suitable tunnel location according to their specific requirements concerning depth and soil cementation in liquefied soils.
Manuscript profile
