Anomaly Detection in Centrifugal Pumps Using Model Based Approach

• Main Authors: Kumar, Swetha R, U S, Iniyal, V, Harshitha, M, Abinaya, J, Janani, D, Jayaprasanth
• Format: Conference Proceeding
• Language: English
• Subjects: Analytical redundancy; Automation; Centrifugal pumps; Computational modeling; Estimation; Fault detection; Fault diagnosis; Industries; Pumps; Residual generation; System identification
• ISSN: 2575-7288
• DOI: 10.1109/ICACCS54159.2022.9785349

In today's society, technological breakthroughs have led to the automation of most industries in order to ensure smooth operations and minimize human interaction. The automation has led to challenges in an important sector called Fault detection and diagnosis. Rather than univariate alerts, fault detection and diagnostics focus on abnormal situations, which is crucial for maintaining optimum operating conditions and anticipating chemical process hazards. Model-based and process history-based fault detection and diagnosis are the two types of fault detection and diagnosis, with the second being more commonly referred to as data-based. Model-based methods are often based on a fundamental grasp of the process's physics. Using a model-based approach, this paper tries to estimate and identify the types of problems in centrifugal pumps. The pump dynamics are estimated using a system identification approach in this paper. The outcomes of the model estimations are then summarized. Estimated and real values are compared while creating residues. Studies are carried out to look into the consequences in both faultless and faulty situations.