Direct- Fully Coupled Solver [Strange Behavior]

Dear All

I´m working on a Natural Convection Tower model. It has the Heat Transfer and Low Re K-e CFD interfaces. And I am using Time Dependent, Direct Fully Coupled method.

I undestand that reciprocal of step size is the indicator of convergence velocity, and we all wish it to be very small. The issue I have been observating, is that every once I pause the computing process to check on the actual results on the heat transfer and flow parameters. And then I continue with the calculation, convergence of several time steps accelerate and then the reciprocal time step rises again and the convergence get stuck again. Why is this happening, every time I pause the calculation, and then continue it, time and the modeling go foreward to an aparent local time convergence of the variables of the flow . This only for a few time, then reciprocal indicator go up again. Can I rely on the results of a problem solved by this procedure?? ( pause, continue). Is there a numerical reason, related with the algorithms of the solvers wich can explain this solver behaviour.


Thank you all.