Complex industrial process has the characteristics of uncertainty, nonlinear, non-Gaussian, large delay and strong coupling, so it is difficult to build linear control model. Particle filter (PF) algorithm can be used in global optimization of nonlinear, non-Gaussian system, making the model reflect the real system state accurately. This paper proposed a partical filter based radial basis function (RBF) neural network method, and applies it to the study of synthetic ammonia decarbornization production process. The synthetic ammonia decarbornization process is a complex industrial production process, whose on-line control model is difficult to establish. Some simulation study with the synthetic ammonia decarbornization has shown that after using PF it has better performance than using only fuzzy neural network. The result also shows that the system is more effectively controlled after using PF algorithm. It provides an efficient way for the complex system modelling and optimization control research.