Alum Plate Bend Test Fail 2

[ElronMcK]

Amateur question. I have been bending my practice weld coupons around a small pipe, then bending tighter in a vice - just to check ductility.
Photos below are pre-bend and post-bend
Background:
Uphill, 3G
One-sided complete joint pen at butt joint (no gap)
6061, 1/8" thick
3/32" 5356 filler (roughly 2" of filler per 1" of weld)
3/32" tungsten with 16 to 17 cfh Argon, gas lens
About 115 amps, 75% electrode negative on Syncrowave
Most welds had full burnthru
No pulsing, relatively steady foot pedal, moderate SS wire brush cleaning prior
Coupons cut to approx 1" wide strips x 5" long
Some welds show very high ductility, but occasionally a weld fractures right at the heat affected zone ("HAZ"), without too much tension straining. Brittle.

I keep preparing to take my 3G test for AWS D1.2, but this is humbling. I would like to determine what are possible contributors. I'm not just the observer - I did the welding - very little difference between the welds - from prep to last puddle.
Any advice is welcome - but especially guys with lots of aluminum experience.

 

[david339933]

My apologies, I didn't see D1.2 I was referring to CSA W47.2. I am surprised there is a difference.

 

[weldstan]

ElronMcK's photos of aluminum bend test specimens is how we prepared specimens for nearly 20 years after some initially poorly prepared specimens resulted in failures both in welder qualification and procedure qualification.

 

[david339933]

If I had samples bent this way, it would be an automatic failure. Uniform bending is required.

 

[ElronMcK]

Okay, cut me some slack. Yes, the bend radius is non-uniform. I'm practicing for the test, and bent those coupons in my garage with my own $17 home-made bender. But zero cracks, guys.
So much for getting an "attaboy" from the group. Tough crowd.