Learning from Reality: Lessons from Centrifuge Models

Abstract

Time does not go round twice. Although we may expend great efforts in forensic engineering to determine the possible causes of disasters after they have occurred, we will never know the precise antecedents. But we should also ask ourselves how we can learn from failures in order that engineers in future may avoid similar problems. Centrifuge models can recreate geotechnical reality but at small scale, and in which the density and stress history of every piece of soil can be established. Models can be fully monitored as they fail. And model events are also practically repeatable except for controlled variations, so we can learn scientifically. Examples are given of slope and excavation failures. Lessons will be drawn which are relevant for the prevention of flowslides such as in Hong Kong at Sau Mau Ping, and the prevention of deep excavation failures such as occurred at Nicoll Highway in Singapore.