This is a very complex issue: in interval analysis we use a lot of heuristics but it is quite difficult to determine what is the right combination that will be the most efficient to solve a problem. And choosing the right parameters for these heuristics may have a very large influence on the computation time (and we mean a really large with a decrease factor of the computation time that may be ).
Usually it is good policy to use filters that may be generated with simplification procedures (see section 4.2). For example the 2B-consistency, see the Hull_Consistency procedure, section 4.2.1) with a repeat factor that is about the diameter of the initial range divided by 1000. Similarly it is good policy to use the 3B method (see section 4.5) with an `ALIAS/Delta3B` value similar to the factor. You may also consider looking at the ALIAS-C++ library that provide specific filters (not all of them are incorporated in ALIAS-Maple) that you may use to write your own specific simplification procedure.