Application of principal component analysis method (PCA) for fault detection in chemical plants

• Autor(a) principal: Mendes, Thessa Fuzaro
• Data de Publicação: 2020
• Outros Autores: Souza, Davi Leonardo de
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/6335
• topic: Supervisión; Detección de fallas; PCA.; Monitoring; Fault detection; PCA.; Monitoramento; Detecção de falhas; PCA.

Control systems are used in chemical industries to reduce the value of process variable deviations from the desired value known as setpoint. Even if conventional controllers contribute to reduce those errors, there is a possibility to occur system faults, which are a not allowed deviation due to some characteristic property or system parameters. Developing new fault detection techniques is the key to meet a demand growing complexity of industrial systems and their performances that aim to achieve better efficiency. This work aims to apply the Principal Component Analysis (PCA) method to detect faults in chemical plants. PCA collects historical process data and constructs a statistical model from them, besides allowing the order reduction of multivariable models to facilitate its implementation. Two case studies were performed involving CSTR (Continuously Stirred Tank Reactor) with heating jacket and a non-isothermic CSTR in order to verify the efficiency of the proposed method in detecting failures in monitored control systems. Both failures in sensors and systems submitted to step disturbances were assessed using PCA and T2 of Hotelling and Q statistics. The PCA proved to be an efficient method in fault detections involving the case studies presented, which indicates its potential to be applied in chemical industry controllers.