Hourglass Control -- ULSPH
Shuaihao-Zhang opened this issue · comments
For elastic case, it seems that we can not use Riemann solver for density. The result is not momentum conservative if we use Riemann solver for density. Below shows the result by using different combinations of Riemann solver for density and pressure.
(1) Integration1stHalf
and Integration2ndHalfDissipativeRiemann
By using this, we can get correct results.
The density and pressure distribution are also smooth.
(2) Integration1stHalfRiemann
and Integration2ndHalfDissipativeRiemann
The result is not momentum conservative.
(2) Integration1stHalfRiemann
and Integration2ndHalfRiemann
The result is not momentum conservative.
For plastic case, using Riemann solver for the density can make the result more stable.
For the Taylor bar case, if we use:
(1) Integration1stHalf
and Integration2ndHalfDissipativeRiemann
We can see that, those particles located at the edges of the column is not uniform.
But if we use:
(2) Integration1stHalfRiemann
and Integration2ndHalfDissipativeRiemann
We can see those particles are uniform.
For the pressure, we can only use dissipative Riemann, otherwise, if we use:
(3) Integration1stHalfRiemann
and Integration2ndHalfRiemann
The simulation has crashed.
Please also check possible reason for these results.
I checked the values of parameters when with and without Riemann solver for density.
I found that, when with Riemann solver for density, the density and pressure distribution are more uniform, and the deviation between the maximum and minimum values is smaller, as shown below:
Also, when with Riemann solver for density, the force caused by pressure is larger, and the shear force is smaller, as shown below:
I think the shear force is to oppose the motion. The shear force for the simulation with Integration1stHalfRiemann
is smaller, so the beam can not stop and go back as the simulation with Integration1stHalfNoRiemann
, as shown below:
But I haven't found the reason why the shear force is smaller when using Integration1stHalfRiemann
.
@Shuaihao-Zhang OK. We will not use Integration1stHalfRiemann since it underestimates tension too much,
A design decision is made.