Applications of sub–period division strategies on the fault diagnosis with MPCA for the biological wastewater treatment process of paper mill

Being a widely used technology in papermaking industry, the fault diagnosis of the sequence batch reactor (SBR) wastewater treatment process has been a significant challenge owing to the batch characteristics. In order to decrease the complexity of monitoring the SBR process, the inherent multi–period characteristics has been considered in this study. The conventional multi– way principal component analysis (MPCA) method has been improved with sub–period division strategies (Sub–MPCA) to diagnose the faults in the SBR process. Aiming at identify the most applicative sub–period division strategy for the SBR process, four types of strategies (Scenarios 1–4) have been tested and compared. Beginning with the off–line modeling, the training set data was used to motivate the sub–MPCA models and acquire the control limits of T2 and SPE statistics. Subsequently, the fault alarm rates (FARs) of the developed models were estimated to verify the models. Finally, the fault diagnosis performances of the models were evaluated with the testing data set from an abnormal batch. After the examinations of the four sub–period division strategies in SBR process, the result revealed that a multi–sub–period algorithm based on the similarities of the loading matrices between the adjacent time slices (Scenario 3) demonstrated the best performance with fewer diagnostic error, which was identified the most accurate model for the fault diagnosis in the SBR wastewater treatment process.