Skip to content. | Skip to navigation

Unusual engineering

Unusual engineeringWhy would you build a bridge halfway across a river? And how do you build a circuit without designing it first? Read on to find out.

The Øresund Bridge

If you want to cross the Øresund Strait from Sweden to Denmark, you take the Øresund Bridge. The crossing is about 16 kilometres - but the bridge is only half that length. Halfway along, drivers plunge under the water.

That's because the designers of the bridge faced an unusual problem. They had to find a way to let ships pass the bridge, something that would normally be easy: you simply raise up a section of the bridge so the ships can pass underneath. But the nearby Copenhagen International Airport made this impossible, as raising the bridge would cause trouble for passing planes. So instead of going up, they went down, converting the bridge into a tunnel for the last part of the crossing.

This meant building an artificial island, Peberholm, halfway across the water. As a bonus, the island is the site of a scientific experiment: nobody is allowed to visit it, except for one yearly trip by biologists, who are monitoring the ecosystem there. The idea is to see how quickly and effectively nature will colonize it without any input from humans.

As if all this wasn't difficult enough, a nasty surprise was waiting when construction began: 16 unexploded World War II bombs, lying on the seabed along the route. Despite this, the bridge was finished three months before it was due.

Evolving circuits

Designing an electronic circuit can be a painstaking task, but some circuits don't require design at all.

The idea of creating hardware through evolution instead of design started with Adrian Thompson in 1996. He took a chip called a "field-programmable gate array", which can be completely reprogrammed to act as almost any electronic device. He started with 50 completely random circuits, and used a computer to measure which of them was best at telling the difference between two sounds. The worst ones were ditched, and the best ones were "bred" together, creating new circuits that were a mix of the old ones – with the occasional random "mutation" thrown in.

Of course, none of the starting circuits came close to succeeding. But Thompson kept repeating the process, and slowly they improved, until – after around 4,000 generations – the chip worked. But the strangest part came when he looked at the final design the process had come up with. It used only 37 of the 100 programmable parts of the chip, far fewer than any human-designed circuit would have. Stranger still, some of them weren't even connected to the main circuit, and yet if they were turned off, the whole thing stopped working.

The circuit seemed to be making use of effects like the magnetic field produced by flowing electricity, something that would have been almost impossible for an engineer to achieve on purpose. Apparently, evolution can sometimes solve engineering problems better than people can – but would you rely on a circuit if you didn't know how it worked?

Related links