DATA-DRIVEN PILE OPTIMIZATION UTILIZING COMPUTATIONAL DESIGN

As part of a major design-and-build pharmaceutical project in Denmark, the foundation design for 6 large production and storage facility buildings was required. The foundations for these structures were subject to substantial vertical and lateral loads. Preliminary analyses identified piled foundations as the preferred foundation design concept. Over 200 load combinations were given for each of the principal and auxiliary support resulting in more than 20 000 individual calculations for some buildings. To address this number of load combinations and foundations necessitated, a strategic simplification into peak and lower bound values or leveraging computational optimization and this way calculating all the load cases.

A specialized automation algorithm was developed for individual pile group design assessments, employing structure-specific soil profiles, geometrical restrictions and prescribed point and surface load conditions for each column.

During the construction phase, pile load testing identified that the realizable pile load-bearing capabilities was greater than the predicted values. This prompted an extensive re-evaluation of the pile design for later stages of the project. Outcomes from the re-analyses identified a potential reduction in material use by up to 30%. The collected pile testing data were integrated into a bespoke machine learning model predicated on measured pile capacity values.

The application of this machine learning model across the remaining buildings facilitated an expeditious and efficient reassessment of newly projected load conditions. This innovative approach combined empirical testing with advanced analytical models to enhance the overall efficacy, sustainability and economy of the foundational designs.