PVC polymerization MICA monitoring and fault diagnosis based on OFA

The multivariable statistical model of batch process cannot be built directly because of its asynchronous batches. Moreover, the sample data of the model do not always fall into the region of normal distribution ideally. To solve this problem, all of the batches are projected onto an orthonormal function to get the approximate coefficients to make them synchronous within every variable. The coefficients drawn from historical trajectories with orthonormal function approximation (OFA) are utilized to build the multiway independent component analysis (MICA) model, and the corresponding ones from tested trajectories are projected onto the model for process monitoring then faulty variables are diagnosed on contribution plot. The results of simulation from PVC polymerization process show that MICA approach based on OFA can discover the abnormal of process correctly and discriminate the faulty variables clearly.