Thanks Mark. That’s a pretty useful method.
I think I’ve cracked the code (so to speak). The logic appears to follow the same methods adopted for commercial finite element programs. Below is a snippet of the local axis definition from ETABS’s analysis manual:
To summarize, the local X axis is decided by the direction of the curve (start point to end point), the local Y is parallel to the Global Z axis and once the local X-Y plane is known, the curve normal direction can be found from the right-hand-thumb rule.
I’ve tested this out on a few examples and it seems to hold. I’ll leave this open in case I’m wrong and someone can advise on the correct approach.