Mathematical disasters

The millenium bridge was not simply a case of a bridge designed in 2-d instead of 3-d (it would have had to be designed in 3-d to account for wind loading). It was a math error, but it was much more complicated - essentially, the slight sideways motion people imparted to the bridge while walking caused the bridge to sway slightly, which in turn caused the people to walk in step with it, creating a positive feedback loop.

The article describes the superconducting supercollider as a forerunner to the LHC, however the SSC was designed to produce collisions with over 3x the energy of the LHC.

Also, I'm surprised the Ariane 5's first launch didn't make this list: http://en.wikipedia.org/wiki/Ariane_5_Flight_501

Inertial guidance code and systems were reused from the Ariane 4, but the greater thrust of the Ariane 5 caused a 16-bit signed integer value to overflow resulting in a cascade of problems culminating in the loss of attitude control and destruction of the launch vehicle.