Hi Marcom & CS2011,
More efficient than what?
This is probably a google translate issue or language issue
I think he means `can you` and the answer is more complex:
You can't twist a centre piece 90 deg without affecting other pieces: A 90 deg turn changes the parity of the puzzle so you would get 2 edges needing to be swapped, etc. So if you turn one centre by 90 deg, you need to turn another centre by 90 deg to correct the parity. If one centre has no orientation and the others do, it would be possible to have this `hidden` (depends on the order you are solving the cube, the method or if it has arrows on all faces).
EDIT: Maybe I'm making this too complicated:
I usually solve a supercube by taking into account the orientation of centres proactively, and using sequences that don't disrupt them. This for me is a philosophical thing, because I believe that is the challenge of the puzzle, rather than adding a fix in hindsight.
RU' RU RU RU' R'U' R2 will turn the U centre (U') and the R centre (R) and not disrupt many pieces. This might be what you are after because I assume you want to fix side centres as part of a layers method? Combined with Marcom's pure algo for U, I think it is OK?