This is the first part in a series of posts related to debugging convergence. This post talks about large sliding problems in particular.
Sliding contact is an imperative characteristic defining the functionality of many products. In this article, I aim to help in debugging convergence issues while modeling and simulating large sliding contact problems in ANSYS Mechanical.
A conventional practice for modeling contacts in a simulation environment tends to accommodate slight inter-penetration of mating parts in order to allow the solution to converge. Inter-penetration can pose serious concerns in the form of under prediction of force reactions and stresses for a sliding contact scenario. The lack of solution accuracy is in the name of solution convergence. In order to derive a reasonably accurate solution for a sliding contact scenario, we should strive to regulate contact penetration to a bare minimum.
Two Approaches to Sliding Contact Problems
Let’s get into the nitty-gritty of solving sliding contact problems. Now ANSYS Mechanical’s settings for penalty-based methods (Pure Penalty and Augmented Lagrange) allow for some penetration (depends upon contact stiffness) leading to easier convergence. Results are not accurate with the penalty-based method. Despite this, many chose to use this approach in order to achieve solution convergence.
Normal Lagrange formulation guarantees almost zero penetration, with good solution accuracy, because there is no contact stiffness in the normal direction. Instead, the method uses some additional contact degrees of freedom i.e. contact pressure acting normal to contacting surface in order to prevent penetration and a tangential contact stiffness based on penalty method.
So the Normal Lagrange formulation can handle large frictional sliding problems more effectively. It is not suited for sticking application, i.e. valid only for frictional/friction-less contacts. Conversely, this method can be used where penetration is undesirable – as in applications such as snap fit, gears & other sensitive applications where penetration leads to less accuracy in the results.
However Normal Lagrange formulation is not the proverbial knight in shining armor for these applications.
Solution has not converged after 12 iterations for contact status change!
Achieving Solution Convergence with Normal Lagrange Formulation
While working with the Normal Lagrange formulation, many of you would have faced this challenge. In addition to it be very frustrating, the time consumed to achieve solution convergence reduces our engineering productivity.
Typically when we activate Normal Lagrange formulation, the ANSYS solver, by default, bisects at the 12th iteration due to contact status change even though the force convergence trend is good. The figure below illustrates this. If the bisection were not to happen, the solution were likely to converge in the next iterations.
Can we increase this bisection limit? Yes! This, little known, undocumented key is available with the CUTCONTROL command.
Command Syntax: CUTCONTROL, NCSI, VALUE
In ANSYS Workbench, this command can be inserted in the tree under analysis system as shown in the below image.
Seen below is the force convergence behavior of a demo case study with and without using the CUTCONTROL command.
Generally, in industry, there is a misconception that Normal Lagrange is not preferable for achieving convergence in many cases. As demonstrated, this contact formulation is best suited for large sliding problems which is both, accurate and faster.
If you encounter problems with large sliding contacts, please do try my suggestion and let me know your feedback. If you have a better solution in mind, please do share in the comments section.