Abstract: An improved BP algorithm based on PSO is proposed.The algorithm in the network learning process using particle swarm algorithm global search, nonlinear inertial coefficient to find the optimal weights, then back to the part of the particle parameter assignment, so as to avoid premature convergence, further improve the original particle swarm optimization algorithm. The BP neural network model is constructed based on the dimensional nonlinear function.The simulation results show that the improved algorithm has good fitting ability to two dimensional nonlinear function and fitting error is improved.