Fillet size

Rather than go further off topic in the thread which brought this to my attention, I thought I would start another thread. I've just been reading Keith Bontrager's article in Bike Tech Volume 4 Number 2 about what fillet sizes lead to the strongest joint.

When I learnt how to braze I was taught that the ideal fillet depth in the middle was at least 4 times the tube thickness, based on the UTS of the brass and the principle (as I understand it) that the strength of the joining material should be at least equal to the strength of the joined material:
Attachment 89539
This makes sense to me (and obviously to many others) from a simple material properties point of view.

The interesting conclusion of Bontrager's article is that he claims that of the fillet sizes he tested, the smallest size (radius / depth at deepest point of 0.10in/2.54mm) should be the strongest. It's a small sample size but I don't think it's an unfair conclusion. According to the 4x rule of thumb this would only be enough for extremely wispy tubing (0.635mm wall thickness... has there ever been a tube this thin at the butts?). Bontrager is using 531 (0.035in/0.9mm) and Ishiwata CrMo (0.040in/1mm) tubing, which according to the rule of thumb would call for a fillet more like 0.15in deep. In practice the difference between these two measurements is probably too small to worry about, but the 4x rule of thumb was framed, to me at least, as a minimum. Bontrager shows a pretty significant dropoff in strength as he increases fillet size up to the maximum of 0.75in/19mm. So I have a few questions to throw out there to set my mind at rest:

1. Is KB's direct inference of tensile strength from hardness an accepted assumption? If you hadn't noticed already I'm not an engineer but the idea that higher hardness necessarily = higher strength stuck out at me. Towards the end he does caution against embrittlement in the very small HAZ caused by TIG welding. Intuitively, and based on what I've heard before, I would have expected that a larger HAZ with a gentler gradient would be less prone to fatigue failure as it would cause less brittleness in an area of high stress.

2. Per 1, here annealing is equated with getting weaker, and hardening with getting stronger. If it is this straightforward, why do we air-cool brazed joints rather than trying to artificially cool them to get nearer the cooling rate of TIG (I'm thinking of adding heatsinks after brazing and using moving air to cool the joint evenly rather than quenching, obviously).

3. Given that even with a very small flame it isn't possible to lay a fillet much smaller than what KB has tested, what are the downsides of trying to go as small as you can (disregarding aesthetics and ease of finishing)? He only warns against using rigid fixtures which may cause cracks in the heated brass due to non-uniform expansion.

4. Are KB's conclusions applicable to all steels now used for framebuilding, given that their alloying and heat treatments should make them react quite differently to heat?

5. (the acid test!) Has anyone ever noted any real life joint failures associated with the size (large or small) of an otherwise flawless fillet?

Cheers,
Ben

Attachment 89590Attachment 89591

Black bike rode from the Straights of the Bosphorus to Tienamen Square & the blue one was raced for years before roof rack incidents.

36mm HT on the blue, 37mm OD on the black, both 1.125" steerer, do the math for wall thickness.



- Garro.

Nice. Thanks Steve, that's food for thought. Looks like the fillets between TT and DT on the cut-up one are less than 1x thickness of the thinner tubes...

The bikes did not, I hope, belong to the same person.