Research and application of the distillation column process fault prediction based on the improved KPCA

With the development of modern intelligent, automation and integration of the increasingly complex industrial process control system, the traditional prediction methods of faults perform not well, so it has faced a huge challenge. In this paper, a new improved kernel principal component analysis method is presented which uses the concept of indiscernibility and eigenvector applied to distillation column process fault prediction. Compared with traditional statistical techniques, improved KPCA not only can remove variables with little or no correlation with the fault, but also can reduce the amount of datas calculated by K. Applying this new method to distillation column process fault prediction, the simulation results show that the proposed methods have great advantages. Compared with the traditional KPCA, the improved KPCA improves the ability to predict the process failure caused by small disturbance and becomes more effective.