2 bugs found

wlodek29

I have found 2 bugs/incorrect behaviours.

The first one is related to particle distribution and is rather algorithm's fault then a real bug. If you take CASEFLOATING from example folder, and change box to cylinder by replacing

setmkbound mk="50" /
drawbox
boxfill solid /boxfill
point x="10" y="2" z="3" /
size x="2" y="2" z="2" /
/drawbox

by

setmkbound mk="50" /
drawcylinder radius="1" mask="0"
point x="11" y="3" z="3" /
point x="11" y="3" z="5" /
/drawcylinder

you will find that cylinder starts to leak. It happens when the wave reaches the cylinder, which is approximately after 5s from the beginning of the simulation. The lattice is originally set to 2, if you set lattice to 1 cylinder starts to leak immediately.

The second one is a pure bug in GenCase. If you try to set manually mass and center of a floating object, center will be set to 0. If you remove mass assignment, center will be set correctly.

SLongshaw

Regarding your first bug, this is in fact more of a general issue with the boundary condition used in DualSPHysics. Simply put, the speed of sound and CFL conditions (as well as the dp value) are too low in the example to cope with the curved surface of a cylinder, so particles pass through. This is highlighted by the fact that in increasing the spread of particles by using 2 per point, you decrease the effect.

To improve this you can: decrease the CFL value, increase the value of Cs0 and also increase the number of faces used to draw the cylinder (take a look at the GenCase PDF to find the sphere and cylinder specific drawing options) as this will increase the number of underlying polygons used to create the cylinder, which in turn will improve the positioning of points, and of course decrease the dp value (more particles = less room for numerical error in the boundary).

Regarding your second point, I'm not so sure, this is something the guys who develop GenCase will have to respond to.