Discussion Forum

Dear COMSOLers,
we are facing the following issue with the Heat Transfer Module.
We are carrying out a simulation campaign in order to evaluate the temperature pattern and thermal behavior of a standard-shaped object. The temperature range concerned includes a phase change for the aforementioned material, from solid to liquid. As you well know, the phase change can be modelled through a particularly-shaped Cp (specific heat capacity - see Fig. 1). The preliminary results of the simulations showed temperature values outside the phisically-relistic range.
In order to better understand where the issue comes from we tried setting up a simpler case, where the component was just a 1D segment. Please find attached the MPH file. We defined the material using an anlytical function approximating the coefficients, we set an uniform initial condition at -20 °C, isolated borders, and a gaussian heat source defined in the whole domain. We performed a time domain analysis up to 1000 s, with a Strict time stepping with maximum time step of 50 ms.
We compared the COMSOL simulation results with a Matlab script implementing the Crank-Nicolson FDTD method with the same parameters, same mesh definition and same conditions. Please find attached (Fig. 2) a plot of the difference between the final temperature values as obtained by COMSOL and by the Matlab script. As you can see, in correspondance of the phase change the error presents an unexpected behavior.
Although the magnitude of the error may be not that worrying, we suspect that this behavior may detrimentally affect the 3D simulations.

Does anybody have an opinion about this? Has anybody already experienced something similar?

Thanks in advance for the support, waiting for a kind reply
Giovanni and Andrea