Hello, I am trying to solve a phase change problem. With the specified heat conditions (a heat flux and convection), the phase change material melts, and convection also starts to take effect. I am trying to model this by including a laminar flow physics and coupling the heat transfer with fluid flow as non-isothermal flow. The solver partially solves and then gives an error message: "Nonlinear solver did not converge. Maximum number of Newton iterations reached. Time : 197.40293977148437 Last time step is not converged." or "Repeated test failures, may have reached singularity".
I am using COMSOL 5.3a, and the specify density button in the coupling is set to "customized linear density", as the usual "density from heat transfer interface" gives an error saying "density reference not defined". I have been stuck with this problem for about two weeks now, and any help with the coupling or solver settings would be appreciated. The model is attached.
Regards Sai Phani Kiran
I can't open your file because mine is 5.2a. But, there are many issues you might want to look at:
1. It could be just because the solver needs more iteration step. All you need is just to continue your run or increase the iteration step limit.
2. Your mesh is not appropriate.
3. Numerical instability that need you to set inconsistent stabilization
For density part, I feel that reference pressure and temperature probably haven't been set yet.
Hi, considering your situation i can suggest you to check the comments below:
-I can't realize your convergence graph because i have the same issue as Mr. Shyu has. However if your graph tends to converge to your predetermined criteria, you can increase your maximum Newton iteration number following the way; Study-Solver Configurations-Stationary/Time Dependent Solver (I don't know which study type you use)- Segregated 1-Maximum number of iterations.
-You can define the velocity field from laminar flow to the heat transfer as a model input under the heat transfer in fluids by changing the velocity field from user defined to velocity field (spf1). But you have to delete the non-isothermal multiphysics first.
-You can check the user guide and you will see when working on liquid metals you have to choose extended Kays-Crawford model for your model and you should determine the density from the solid side (you can still define the variable density values under the material data) by doing this Comsol automatically gets the value from the material.
I hope it helps you.
When solving phase-change problems it might be useful to reformulate the heat-transfer equations, e.g. using enthalpy-kirchhoff method. In that case the numerical solution procedure is more stable ...
- Badges
- Science topic