In this talk, we concern a rank constrained semidefinite programming (SDP) with full nonnegative constraints on the matrix variable. As an important application, a class of combinatorial optimization problems can be equivalently reformulated to that rank constrained SDP. However, it is, in general, a nonconvex and difficult problem due to the rank constraint. Based on the Clarke exact penalization principle, we design a difference of convex functions approach (DCA) to solve the rank constrained SDP problem and prove the same optimal solution between the rank constrained problem and the DC problem with suitable parameters. The subproblems of the DC problem are solved by a semi-proximal augmented Lagrangian method (sPALM). If the DC parameter is large enough, we show that the sequence generated by the DCA always satisfies the rank constraint in the original problem. Numerical examples demonstrate that the proposed approach is very efficient for obtaining the feasible solution of the rank constrained SDP.