Complex industrial processes possess several critical features, such as uncertainty, nonlinearity, and large delay, which present significant challenges to the construction of real-time control models. This paper proposes a particle filter-based radial basis function (RBF) neural network to model and control complex industrial processes. The proposed method employs the particle filter technique for estimating the system's prior information to improve the RBF neural network's learning speed and expression capability, hence making real-time control possible with satisfactory static and dynamic performances. The proposed modeling method is applied to a real-life synthetic ammonia decarbonization process for performance evaluation. The simulation and experimental results illustrate that the proposed neural network system steadily refines the parameters as this real-life process proceeds and achieves a higher level of modeling accuracy than an existing method using a fuzzy neural network. The proposed method provides an effective approach to model and control similar complex industrial processes.