Time-dependent Heat Transfer study - strange periodicity

Dear COMSOL users,

I'm modelling a (reasonably simple) geometry in which I have a pulsed heat source with a period of 3 minutes (1 minute the source is on and 2 it is off). The geometry is a cylinder with different layers (of different materials), one of which is the heat source. All the surfaces are cooled by natural convective cooling in air and irradiation.

I've been running a quite long calculation to see how long it takes to reach the steady-state temperature, but in my results, for the temperature on the surface, I see a strange pattern: the temperature seems to fall every 2000s and then goes up again (see attached file).

Have you ever experienced it? Is it something physical or maybe something I'm doing wrong in the calculation?

Thanks for your help