On July 18, Roscosmos held a press conference and issued a release officially explaining what had already filtered out through unofficial sources: the July 1 Proton-M launch failure occurred due to three DUS angular velocity sensors having been installed upside down.
“Installing these devices is complicated and awkward work,” said deputy head of Roscosmos, Alexander Lopatin. Lopatin went on to explain that either the worker who installed the devices or the engineer who wrote the blueprints could have been at fault.
A few days earlier, investigators from the state commission conducted an experiment to assess whether it was possible to install the DUS upside down. They found that it was more difficult than the correct installation and required special tools to make the instruments fit without the pins designed to help the technician align it properly, leaving noticeable damage on the sensor plate. Upon review, it turned out that the DUS on the crashed proton displayed similar damage markings.
Once installed in the rotated position, there was little opportunity to discover the error. Color coded cables worked just as well in the inverted position. Although there were arrows on the DUS to indicate the upwards direction, there was no corresponding arrow in the mounting plate as a reference point. The installing technician’s supervisor and quality specialist overseeing the work signed off on the procedure, although it is unknown to what extent they evaluated the hardware.
One likely outcome of the investigation, according to Lopatin, will be to require photo and video documentation throughout the rocket assembly process. This is a step that was recently initiated for the Breeze-M upper stage in response to repeated failures of that vehicle.
The commission is still investigating some aspects of the crash, such as the slightly early liftoff that dominated initial speculation; however, there seems to be a consensus that the upside down DUS were the underlying cause of the launch failure.
The Proton-M may be back to regular launches in September, although at this point it is practically impossible to make up all the year’s originally scheduled missions.
Below, live video of the original launch failure: