Sousa’s work — which was supported by FCT (Fundação para a Ciência e Tecnologia), in part by the MIT Portugal Program, and which culminated in her successful dissertation defense in December 2009 — led to the compiling of what could be the most comprehensive database of tunnel accidents to date, and the development of a novel methodology for assessing tunnel project risks.
In particular, Sousa’s model predicts changes in geology that could steer designers and builders toward safer materials and construction methods for tunnels (including for high-speed rail systems, an important area of MIT Portugal’s Transportation Systems research).
Starting the research
Sousa, a native of Lisbon, came to study at MIT after meeting CEE Professor Herbert H. Einstein at an engineering conference in Lisbon. Einstein, a member of the MIT Portugal Transportation Systems faculty, became her PhD advisor, and CEE Professor Daniele Veneziano, also on the MIT Portugal faculty, became her co-advisor.
Sousa’s decision to focus on tunneling accidents was a natural continuation of her undergraduate and master’s degree studies in civil engineering, as well as work she did for an engineering company belonging to the Lisbon Metro that supervises construction and design.
“Although most tunneling projects have been completed safely,” she says, “there have been several accidents during the last several years that have led to delays and higher costs — even loss of life. My idea was to look at existing knowledge of how these accidents occurred and build a system that would allow you to predict geology and choose the best construction methods.”
Database and methodology development
Sousa created a database of 204 tunnel construction accidents that have occurred worldwide to systematically assess the risks associated with tunnel construction (she was a bit surprised, but pleased, to get details from some of the principals, as well as from public accounts). She then analyzed the data and created “influence diagrams” showing key factors and the interactions between them.
Sousa then developed her risk-assessment methodology by combining a geologic prediction model (predicting geology before tunnel construction begins) and a decision-support model (one that allows users to choose a construction strategy that indicates the lowest level of risk).
The next step was to apply the methodology to a case study — the Porto Metro accidents. Here the results were powerful: the model accurately predicted a change in the geology in two zones where the tunnel incidents occurred. Using this model, Sousa says, might have suggested a safer construction method — and might have helped prevent the accidents.