General Mechanical

General Mechanical

Convergence issue in Non-linear analysis

    • Shekhar1234


      I am trying to simulate a soft pneumatic actuator in Ansys. The material that I am using is Elastomer (Neo-Hookean).

      The actuator has 4 chambers out of which I am actuating only one i.e. only one chamber has pressure loading as shown in pictres below.

      This figure shows loading at 9.69s when my solution stops solving. My final load is 0.05 MPa at time 10s. I have given 10 load steps linearly increasing manner and as I was having convergence issues, I increased substeps limits in last step to Min 50-Max 100. I also have added command snippet NEQIT,60 to increase number of equilibrium iterations.

      This figure shows the pressure variation with time.

      My mesh looks like:

      I have refined the mesh near the channel I am applying my pressure into.

      The problem is: the solution stops at a time of around 9.7 sec and shows a lot of errors like:

      Attached below are the force and displacement convergence plots and the unconverged solution at time = 9s:


      My questions are:

      1) How to make the solution converge?

      2) If not, can I assume the solution at 9 second as a valid behaviour for a total load of 9/10*0.05 = 0.045 MPa?


      Thanks in advance!

    • Sandeep Medikonda
      Ansys Employee

      Some suggestions/questions:

      • See if the unsymmetric newton-raphson solver (Analysis Settings>Nonlinear Controls) helps?

      • Have you tried a Hex or Hex-dominant mesh? 

      • What is the max. strain you are observing in your model? How does it relate to what you have in engineering data?

      • Plot the Newton-Raphson residuals to understand where the max. residuals are occurring?

      • Use CNVTOL,COMP,, to increase/decrease the volumetric compatibility tolerance (default value is 1e-5).

    • Shekhar1234

      Hi Sandeep,

      1) Max. strain at time 9.4 s is around 1.43 which is outside the limits of shear test data for the elastomer. Maybe I will extend the data points.

      2) Earlier I was using hex-dominant method with the model split into different solid bodies for selective sizing as we need finer elements near the channel. Due to the non-linear bonded contacts between the bodies, it was taking a significantly large amount of time to converge at lower loads and was failing to converge at higher loads such as 0.5 Mpa. That's why I reverted back to using tettrahedral mesh and refined the mesh near the channel instead. It was solving at pretty faster rate (I think it was solving in-core compared to out-of-core in case of hex mesh).

      3) I will try to solve it with unsymmetric NR solver.

      4) NR residuals look like:

      The maximum occurs somewhere near:

      I guess the mesh size is pretty fine there and I have refined it upto 3 levels near there.

      5) The commands you mentioned: In what block do you place that command snippet? Should I increase or decrease the compatibility tolerance?


      Also thanks for your quick response!

Viewing 2 reply threads
  • You must be logged in to reply to this topic.