A new approach is presented to solve nonlinear constrained programming problems (NLCPs) by using particle swarm algorithm (PSO). It neither uses any penalty functions, nor distinguish the feasible solutions and the infeasible solutions including swarm. The new technique treats the NLCPs as a bi-objective optimization problem, one objective is the original objective of NLCPs, and the other is the degree violation of constraints. As we prefer to keep the ratio of infeasible solutions so as to increase the diversity of swarm and avoid the defect of conventional over-penalization, a new fitness function is designed based on the second objective. In order to make the PSO escape from the local optimum easily, we also design a adaptively dynamically changing inertia weight. The numerical experiment shows that the algorithm is effective.