Caio,
Short answer is yes, it takes a long time and don’t remove that line.
For this post’s reference the paper can be found here. As is common with many coupled phase field approaches, a staggered approach is taken here. A simpler problem like this could be solved with a coupled solution - but it would still take a long time.
The best and worst part of the phase field method is that it requires very small steps when a phenomena is active. I.e. in this paper the authors choose inputs (Gc, l, ...) such that the necessary energy to propagate the crack is not reach until a load in the vicinity of t=0.7. This implementation uses load driven displacement load.t=t*u_r so the load is tied directly to the current time. There is no reason to take tiny steps prior to the initiation of crack movement hence:
if t >=0.7: deltaT = 0.0001
Before this time the problem is essentially linear elastic in nature. Once crack propagation starts the solution can drastically change with just small changes in the applied load. I recommend that you go through each line and make sure you are aware of the author’s choices. The reason you stop seeing updates with large time steps is:
err_u = errornorm(unew,uold,norm_type = 'l2',mesh = None) err_phi = errornorm(pnew,pold,norm_type = 'l2',mesh = None)
If you take large steps while the crack is propagating, the method will not converge and in this specific code you will be stuck in an infinite loop of those messages (no max iterations). In my experience with staggered approaches you take large time steps until the solution fails to converge and then refine the timestep at that time. If you changed any of the inputs from this paper, you would likely need to change where the step refinement occurs. Phase field methods can be very stable but at the cost of small timesteps (and no lack of fidgeting). Consider how quickly a crack can initiate and propagate in a real material - and now we want to actually capture it moving in small steps. Hope this helps.