You can set the maximum number of controlled increments as shown below:
I would suggest turning the arc-length off and using 10 increments and the ITER method until you cannot get converged answers for your problem.
There may be other issues with your model also. You are using a large displacement nonlinear solution with RBE2 elements. These rigid elements are linear only and do not have large displacement capability. If the ends of your tube rotate significantly from a rigid body motion standpoint, then these elements cause convergence issues at best.
Also you model the tube with a single solid through the thickness. This does a poor job of capturing bending stiffness and stress through the thickness which I think are important for your problem.