Fault diagnosis of a pH neutralization process using modified RBF neural networks

The fault diagnosis system of a pH neutralization process is developed using the hybrid approach of NeurOn-Line neural networks application with a modified radial basis function (RBF) learning algorithm and G2 real-time knowledge based intelligent expert system building technology. Firstly, a brief description and modeling of the pH neutralization process is presented. Then considering the slow convergence speed of the K-means clustering algorithm of RBF neural networks, a modified K-means clustering algorithm and a self-adaptive adjustment algorithm of learning rate are presented, which obtain the optimum learning rate by adjusting self-adaptive factor to guarantee the stability of the process and to quicken the convergence. Finally, the fault diagnosis system is designed and programmed in detail in the NeurOn-Line with improved K-means clustering algorithm within G2 environment. Normal operation mode and three fault operation modes of the pH neutralization process including pH sensor biased high, pH sensor biased low, and base reagent diluted is simulated and diagnosed. Simulation results demonstrate that these normal and fault operation modes can be quickly and accurately classified. This proves the effectiveness of the whole RBF networks fault diagnosis system and the improved K-means clustering algorithm.