When reliable information is unavailable on the
relationship between properties in a casting and those in a
separately cast test specimen, and where experimentation
would be unfeasible, the size of the test casting should be so
selected as to approximate the thickness of the main or
controlling section of the casting.