For feasibility-check problem, if I use SCA(successive convex approximation) for one of the non-convex constraint, do I have to finish the whole SCA process?

[Mark_L_Stone]

I think SCA is not a reliable approach for what you want to do. it may work, or it may be inconclusive.

To clarify my earlier statement “Given at least one non-convex constraint, the only way to reliably conclude infeasibility is if a global optimizer (or constraint solver) reports infeasibility.”, if a convex relaxation of the original problem is infeasible, then you can conclude the original problem is infeasible; but the converse (or inverse) is not true.