[diggan]

> there is plenty of movement visible on all kinds of bridges

Sure, but "the movements in the cluster began to show an atypical pattern with diverging paths" does highlight that even though movements might be common for all bridges, the particular pattern they discovered for this bridge was uncommon.

Agree that the maximum value one could get from a process like that would be "When will this bridge fail?", but better than nothing, is to get a "This bridge looks like it will fail soon".

Besides, could manual human-driven inspections even be able to give a "more or less exact prediction of when a bridge fails"? If not, then at least a somewhat automated way of getting the same answer would be cheaper for local governments to run, and for the cases where it makes sense, send the human bridge inspectors there.