If you have significant plasticity, a sudden turn around in load can sometimes cause a problem. (I haven't seen this error message before, but I work in APDL.) Try a load of 1199 for your second load step before going to 0 in the next one. Make sure there are a sufficient number of substeps in the first and last load steps to reasonably track the nonlinear behavior. If you still have trouble at the turn-around load step, you can loosen the convergence criteria for that one, then tighten it back up for subsequent steps.
Many thanks for your suggestions. I've made some research and finally I've noticed that it is not necessary to retry the load to zero to achieve plastic strain. Usually, plastic strain is almost the same in the last load step like the step where the load is applied. I could insert 90% of load in the last load step.
You said something about " loosen convergence criterion".
How can I make this? Is there an APDL command to do this?
I have experience only with Workbench.