Free-surface flow past a cylinder, help !
 

I am simulating a free-surface flow past a vertical circular cylinder (Re=10,000, Fr=0.8). The calculation is initialized with the results of a smooth rigid cover (double model without a free surface) in which the vortex shedding has been fully developed. As the wave being generated, the vortex shedding at free surface tends to be slower and weaker towards symmetric structures. However, after a period of time the free-surface elevation at one of the large eddy centers drops very deeply, like a "whirlpool". This singularity eventually diverges the calculation.

Could somebody point out the reason(s) ? Is this a physical or numerical phenomenon ? and how should I deal with it ?

Thanks