Application of fault identification algorithms for sensors based on data reconstruction in fermentation process

A fault identification algorithm of principal component space (PCS) information reconstruction based on T² statistic have been proposed aiming at the defect of difficulty to effectively identify the fault, that qualitative diagnosis can only be implemented using traditional variable contribution rate and the data reconstruction method based on Q statistic ignores the fault information of PCS. The reconstruction value, T² statistic and its control limits are obtained on the basis of defining fault subspace and using the normal process data to calculate reconstructed index and reconstructing the fault data in PCS. In this paper the lincomycin fermentation process is studied and the sensor faults are set and identified by statistical model based on PCA. The results have shown that the method used has good capability in diagnosis and recognize ability.