
ID	0116
Class	KNOWN-LIMITATION
Severity	1
Status	ready

Models: —

Found: 2026-10-17, feature seabed-contact (plan D1 phase 2, B1, base anchors; owner ruling 2026-10-18: the stiction is deferred)

Evidence

include/fhsim_environment/seabed/SeabedLaw.h:41 and SeabedContact.h (AnchorShankLoad, SoilForce): the slip is opposed through the regularised sign u(v) = v / sqrt(v² + v_reg²) with v_reg = 0.01 m/s, so the friction at a slip speed below v_reg is the proportional part of the limit, not a hold. A body at rest on a slope or under a small side load moves with it.
Measured in fhsim_base anchors (ad hoc drops of B1, 30 s, RK45_i and BDF, both finite): FlukeAnchor (steel, shank 2 m x 0.1 m, dropped from 3 m) lands, rolls to a stable pose, then drifts 1 to 4 cm/s laterally and yaws 0.013 rad/s without stopping (bounded, no growth); TorpedoAnchor (4 t, D 0.5 m, L 3 m) sinks about 10 m in 3 s and creeps 7 mm/s. The top-cap part of the TorpedoAnchor creep is fixed by the shank end cap ruling (env 3.2.7, owner 2026-10-18: a cap that faces up gives no force at rest); the remaining creep is the regularised friction.
The anchors are FlukeAnchor and TorpedoAnchor in fhsim_base (src/anchor/); they call SeafloorForcesInterface::AnchorShank and Plate. fhsim_base has no open issue for the creep; this is the one.

Effect

A resting anchor does not stand still: it drifts at centimetres per second and yaws, so a mooring or anchor test that expects a stationary anchor has to accept the drift, and a long run moves the anchor by metres.

Possible fix

A stiction rule: a hold below a break-out force (a stick state, Karnopp's dead band or a Dahl/LuGre bristle) so that friction is static below the limit; or a smaller v_reg for anchors (stability cost under explicit steps). Needs a state, as ENV-0115.

Test that would prove it

A steel FlukeAnchor resting on the bed for 60 s with no load moves less than 1 mm and yaws less than 1e-4 rad/s; the same with a side load below the friction limit.

Risk

State in the law (see ENV-0115); a tighter v_reg makes the friction stiffer, and an explicit step must then be shorter.