The past 60 years have not seen any fundamental change to the design of bridge decks for suspensions bridges – best known in Denmark from the Great Belt Link.

To accommodate the request for ever longer bridges, the Technical University of Denmark (DTU) and COWI, studied how to optimise structures to reduce the weight of the bridge deck, in particular increasing the span. Recently published in the recognised scientific magazine Nature Communications, the results of that research project indicate vast potential.

“We applied different methods for examining how to better utilise materials, which primarily consist of steel and concrete. Initially, we sought to optimise their use in traditional structures by using transverse diaphragms in the bridge deck, thereby achieving a theoretical weight reduction of up to 14 per cent,” says Mads Jacob Baandrup, who carried out the analyses in connection with his PhD project and today works as an engineer in COWI’s bridges department.

With a view to achieving additional savings, the researchers looked at the possibility of altering the structural design. That was done by using topology optimization, a method known in car and aircraft industries, that had not previously been used for large-scale building structures.”In popular terms, it’s about ’emptying’ a bridge girder of its existing elements, providing complete freedom for choosing a new design.

The inner volume of the bridge girder is then divided into a structure of very small voxels (3D pixels), like small dice. The topology optimisation method is then used for determining whether each individual voxel should consist of air or steel material. The result is a bridge girder design that uses the least possible steel without impairing the strength of the structure,” says Associate Professor Niels Aage, DTU Mechanical Engineering, who is one of the world’s leading scientists in large-scale optimization and was responsible for the project analyses.

Specifically, a bridge element measuring 30 x 5 x 75 metres was analysed, divided into two billion voxels, each no bigger than a few centimetres. The result was an incredibly extensive calculation performed by a supercomputer, which would have taken an ordinary computer 155 years to do and is the largest structural optimisation ever carried out.