Autodesk Simulation Mechanical and Multiphysics

[#1]

I want to analyze a steel ball with initial velocity impact the floor.In order to create a nodal point at the center of the sphere, I combine the ball with 8 one eighth ball parts. It's very stranger that the results (Dz of the ball center nodal point) are quite different with different contact conditions or element types setting.

    Initial velocity= -500mm/s  
    impact plane= -10mm (Z direction)
    standard gravity force ( -Z direction)

*design scenario 1 (contact tolerance for impact plane 0.01mm):

 The bounces are tending down, but sometimes the decrease is not obvious between two sequential bounces.

*design scenario 2 (contact tolerance for impact plane 0.01mm; each element is with midside nodes):

 The ball seems to bounce at the same level after the 2nd jump 

*design scenario 3 (contact tolerance for impact plane 0.254mm - default setting of the program):

  The ball will bounce higher and higher.

Q1: How to explain the differences which are described above?
Q2: How to make sure which analysis is correct without experiments? This is a simple model, we can ascertain the numerical result more easily. For a complex model?
Q3: How to define a nodal point with specific coordinate position inner a brick part? (e.g. Import a whole sphere and Autodesk simulation will create a nodal point at the center of the sphere exactly.)

Thanks!               Ccz90

[#2]

There are a few parameters which can influence the solution of drop test via impact planes.  Some of them have dependence on the others;

1. Parameter for (MES) integration method for "General: MES, NLS":   "1" is the default.  It will add high frequency damping (but make the model stable for convergence) and then will make the bounce lower and lower.  And this can be influenced by time step size.  Very small time step size can reduce this damping.  "0" can be used, but might result in worse convergence.  However, "0" is used, to be careful, tiny step size must be used in the contact time zone, otherwise, the model might bounce too high.
2. Time step size:  For the implicit code, especially when simulating impact problems, the time step size can have a huge influence on the results, and the model might bounce too high.  You can use multiple time zones for this.

-Shoubing