Tolerance and geometry control of a deployable/erectable space truss

None of them. Cable tension set by over-centre lock geometry carries 87.2% of the torsion-frequency variance, through an amplification factor EA/T of 450 turning cut length into tension. All dimensional tolerance together carries 13%; the only drawn tolerance in the entire source corpus carries 0.3%. Dimensional tolerance under-predicts the hardware's measured reassembly scatter by about 17×.

Download PDF 8 pp · 152 KB

MODE Structural Test Article4,000 samples + 300 paired-seed cross-checkone solver and a witness

Result

The mechanism is an amplification factor

\frac{EA}{T} \approx 450

that turns a cable cut tolerance into a tension tolerance, so the parameter that matters is a dimensional one that nobody thought to tolerance as such. The single documented tolerance in the source corpus is the opposite case: halving it moves the scatter by 0.1%.

The 17× shortfall is still 10× against the lower confidence bound. The missing source is joint behaviour that appears under load — plausibly a locking sleeve set by finger pressure, for which a 2.7% stiffness variation would suffice. A tolerance budget built from the drawing package would have predicted 6% of what the hardware did.

The two solvers agree at r = +0.999 on paired samples; their only disagreement is a mode-dependent scale on the scatter, predicted to 1–3% by a bias measured before sampling began. 23% of cross-check samples failed to converge, and the survivors were shown to remain representative rather than assumed to be.