You are here
Influence of Penalty stiffness factor on simulation results
Sat, 2009-09-12 07:01 - abeed
The default value of penalty stiffness scale factor in Ansys-Lsdyna is 0.1, when one varies this value, he/she obtains different results, the question, how can one know the right value of penalty stiffness scale factor?
I am simulating the impact of a 6mm - diameter steel solid ball with a rigid falt surface using Ansys-Lsdyna. the speed of the ball at the impact moment is low (from 1 to 10 m/s). specificly, i am intersted in the period between just touching and just leaving the ground for the first impact.
»
- abeed's blog
- Log in or register to post comments
- 22691 reads
Comments
re: penalty stiffness value
Abeed,
There is not a single "right" value for the penalty stiffness factor.
One way to think about this is that you are not really simulating the impact of a ball with a rigid flat surface. What you are doing is to simulate the impact of a ball with a spring on that surface. The penalty stiffness is the spring stiffness. If you change the stiffness of the spring, the results will of course change, because you are looking at a different problem.
As you increase the stiffness of the spring, you will get closer and closer to the "real" problem of the ball impacting the wall. You only approach the real problem in the limit as the penalty stiffness goes to infinity. Of course, practically speaking, this is impossible. Moreover what you will likely find is that, at some level of penalty stiffness, the problem will take much longer to run. This is because the stable time step for the explicit calculation will begin to be affected by the penalty value.
Dear Mr. John E.
Dear Mr. John E. Dolbow,
first of all, Thank you for the guidance. In fact what you mentioned happens when doing contact (which is static) analysis using ansys (not ansys-LsDyna), low stiffness may give inaacurate results and higher contact stiffness gives better results, anyhow low stiffness means quick convergence and high stiffness means slow convergence but high accuracy, therfore as you mentioned, we have to seek the optimum value of the penalty stiffness which produces good results with fastest possible convergence.
this does not happen when i simulate the impact of the steel ball with rigid flate ground using Ansys-Lsdyna explicit dynamics, i started with K = 0.1, 0.2, ...6 the deflection at the ball center was degreasing in accordance with the increasing k with no noticable change in speed of convergence but when i exceed certain k (it is some value around 6) the deflection sudenly jumb to high value (2 to 3 times last before), i do not know the reason, i dreastically reduced the time step but to no avail, please if you know some idea let me know
i used solid164, full-integration, lagrangian, hexahedral brick element for ball and ground. ball is made of steel (E = 196e9 , density = 7800, v = 0.3, Radius = 0.003) the results deviates from Hertz equation awfully. i am using slow impact (Ball speed = -1 m/s) it has to be somhow near to Hertz equation)
reply to abeed
Abeed,
I'm afraid I don't have a good explanation for you with regard to the sudden jump. It could be the result of a whole host of things. I'm simply not familiar with the inner workings of LS DYNA.
You have to keep in mind that many of the explicit contact algorithms in commercial codes are designed to yield something that "works" in the sense that contact conditions are roughly enforced. If you're allowing for some nonlinearity in the bulk response of the ball, inconsistent results as one varies the penalty stiffness are certainly possible.