<![CDATA[Heritage buildings founded on soft alluvial clay settle differentially when the groundwater table fluctuates and adjacent construction proceeds, yet micropile underpinning is still designed deterministically through a global factor of safety. This study quantifies the reliability such practice actually delivers and identifies cost-efficient alternatives. An axisymmetric finite-element model, a hyperbolic load-transfer model, and Mindlin interaction factors predict group capacity and settlement. Five lognormal random variables are propagated by the First-Order Reliability Method and Monte Carlo simulation to obtain reliability indices for the capacity, total settlement, and differential settlement limit states. Kriging and gradient-boosting surrogates trained on a Latin hypercube design replace the group solver within an NSGA-II search over diameter, length, pile number, and spacing. Three alluvial profiles compiled from a 62-sample database are analyzed under an 800 kN column load. Conventional practice (four 200 mm diameter, 12 m long micropiles at a factor of safety of three) returns a system reliability index below one on the two softer profiles, with serviceability rather than capacity governing. Kriging outperforms gradient boosting, reaching test coefficients of determination of 0.88 to 0.93 versus 0.72 to 0.85. The novelty is a reproducible open-source workflow linking a regional soil database directly to reliability-based underpinning design.]]>
Copyrights © 2026
