In 2017 Weiss tested 8 by 8 ft wood shear walls at Washington State University. To simulate decay at the base, she left out every nail below 12 inches. Walls nailed at 6 in on the edges kept 57% of their strength. Walls nailed at 3 in kept only 34%. The buildbot-sheathing programme needs to know why. A robot that misses nails low on a sheet needs to know which misses matter.
Our last two steps were rigid-plastic limit analyses, where every fastener carries its full strength at once. With code-strength stud joints, both models said the decayed 3 in wall should keep 64–80%. They only got near 34% when the 16d nails joining the studs to the sill were made much weaker than the code's design ratio. Milestone 18 left two physical explanations a rigid-plastic model can't represent:
- Slack. The hold-down's anchor rod sits in an oversized hole, so the end studs can slide before the rod bears.
- Peak order. The stiff stud joints may peak, and soften, before the sheathing nails reach their peak.
So today's step builds the same wall as a displacement model and pushes it over to 150 mm. Each sheathing nail follows a tested load-slip curve: 8d nails through 7/16 in OSB, from Shameli Derakshan, Zhou and Ni (2019). The stud joints soften too. Each anchor rod carries nothing until a gap closes. After that it has the NDS dowel stiffness, about 47 kN/mm, up to its design value.
- The model checks out. With elastic-perfectly-plastic springs it reaches the limit analysis's capacity within 0.044% in all 16 cases. Softening alone moves the decayed walls by under a point.
- Slack doesn't matter. A 1/16 in gap, a standard oversize hole, moves the decayed walls by 0.1 point. Even 1/4 in only matters when the stud joints peak early. In that case the 3 in wall still keeps 52–63%.
- Neither does peak order. If the joints peak at a quarter of the nails' slip, the decayed 3 in wall still keeps 71% at code strength.
- Every fit has weak end nails. We ran 960 pushovers. At 0.75 of the code ratio or above, no joint strength came within 10 points of both tested groups. All 30 fits within 5 points have the end nails at a quarter of the NDS ratio. The best keeps 57.8% and 32.5%, against 56.5% and 33.9% tested.
We registered six predictions in the log before running anything. Three held: the limit check, softening alone and slack. Three were refuted: the comparison with milestone 16's single sheet, early-peaking joints and a near-code fit. The milestone 16 comparison was refuted by a modelling convention, not a bug. A rigid stud pinned on its centre line also moves the edge nails, which sit 3/8 in off that line, slightly up or down. That makes it about 1.3% stiffer than milestone 16's parallelogram. With the pins on the nail lines, the two models agree to seven digits.
The result narrows the question. Either 16d end-nailed stud-to-plate joints really carry about a quarter of their design ratio in these walls, or the 34% comes from what a monotonic push leaves out. That means cyclic tear-out at the unsupported bottom edge, or the middle stud splitting, both of which Weiss reports. A tested ultimate and load-slip curve for end-nailed joints would decide the first. A cyclic push would test the second. None of this changes the robot's numbers: an escaped miss still costs 2–3% of a wall on average.
Limits: the push is monotonic, while Weiss's tests were cyclic. The end-nail joint's curve borrows the sheathing nail's shape, with its strength and slip scale swept rather than measured. Studs and sheets are rigid. Code, numbers and figure: research/buildbot-sheathing.