differences between turbulence model
 

Hi guys,
I'm using k-epsilon model for my problem. In residual plot the epsilon term has the most error and it doesn't converge. I changed solution methods and under relaxation factors many times, but it didn't work. What can I do? Is this about that the k-epsilon model is not suitable ?
The case is a submerged intake, the model is VOF and the solver type is pressure-based and steady.