Engineering Judgment, Managing Risk, and the Use of Software-PPT Presentation

PPT Presentation

Pdf Summary

The presentation argues that advances in artificial intelligence and geotechnical software make analysis easier to perform, but also increase the risk of misapplication. AI and analytical models produce outputs based on their data and assumptions; they do not possess engineering judgment. Their results may be precise without accurately representing real conditions.<br /><br />Geotechnical modeling is particularly challenging because subsurface materials are variable, investigations provide limited and discontinuous data, and measurements themselves carry uncertainty. Models simplify complex systems, so engineers must judge whether their assumptions and results are realistic.<br /><br />A Conceptual Site Model (CSM) helps bridge analysis and reality. It synthesizes site information, explains how the site formed, identifies constraints and uncertainties, and provides a reasoned basis for design. Engineers should develop the CSM before building an analytical model, then use the model to test hypotheses and compare predictions with observed conditions.<br /><br />A roadway failure case illustrates the consequences of overlooking this process. An initial slope-stability model did not predict the observed failure. Reducing soil-strength values to force a predicted failure was unreasonable. Review showed that the model used circular failure surfaces inappropriate for the site’s sloping residual shale layer. Using non-circular surfaces with the original, best-estimate soil properties produced results consistent with field behavior and supported a remedial design.<br /><br />The central message is that software cannot replace professional expertise. Sound engineering requires a strong CSM, critical review—especially senior review—and validation of model results against measured and observed conditions.

Keywords

artificial intelligence

geotechnical software

engineering judgment

analytical models

subsurface variability

uncertainty

conceptual site model

slope stability

non-circular failure surfaces

professional expertise