Fuzzy C-Means clustering algorithm is a classical clustering algorithm, which is widely used in data analysis and performs well. However, serious defects in the algorithm limit the application of the algorithm and development. Pharmaceutical process is a very complex integrated system, whose parameters are very complicated accused of the relevance and existence of fault and error. For these issues, the Fuzzy C-Means algorithm is applied to dynamic recurrent fuzzy neural network predictive control. Improve Fuzzy C-Means algorithm with IPSO algorithm to do the clustering of data and failure pattern recognition of system to achieve the early warning and ensure system stability.