Unsupervised Detection of Abnormal Operating Regimes in Industrial Centrifugal Pumps Using Multivariate Condition Monitoring Data
DOI:
https://doi.org/10.69980/bc4agr40Keywords:
centrifugal pumps,, condition monitoring, Isolation ForestAbstract
Industrial centrifugal pumps are critical assets in industrial facilities, where abnormal operating regimes can adversely affect reliability, maintenance efficiency, and process continuity. This study investigated the detection and characterization of abnormal operating regimes in industrial centrifugal pumps using multivariate condition monitoring data and an unsupervised learning framework. A dataset comprising 173,730 observations collected from three centrifugal pumps was analyzed through a workflow that included data preprocessing, sensor-data quality assessment, descriptive statistical analysis, correlation analysis, Isolation Forest-based anomaly detection, and statistical validation. Five operational variables—motor vibration velocity, motor peak acceleration, pump vibration velocity, pump peak acceleration, and outlet pressure—were used for anomaly detection modelling. The results revealed that anomalies were concentrated within specific operating periods rather than being uniformly distributed across the dataset. Pump B on 30 October 2024 exhibited the highest anomaly percentage (59.09%), followed by Pump A on the same date (41.11%), whereas Pump C on 30 October 2024 showed no detected anomalies, indicating a stable inactive operating state. Correlation analysis further demonstrated strong relationships among vibration, acceleration, and pressure measurements, reflecting interconnected mechanical and hydraulic behavior. Anomalous operating conditions were primarily characterized by substantial increases in motor peak acceleration, pump peak acceleration, and outlet pressure. These findings demonstrate that unsupervised anomaly detection can effectively identify abnormal operating regimes in centrifugal pumps without requiring labelled fault information. The proposed framework offers a practical and scalable approach for industrial condition monitoring and supports maintenance decision-making in data-rich operational environments.
References
1. Adaika, H., Tir, Z., Sahraoui, M., & Laadjal, K. (2025). PumpSpectra: An MCSA-Based Platform for Fault Detection in Centrifugal Pump Systems. Sensors, 25(22), 6916.
2. Ali, W., El-Thalji, I., Giljarhus, K. E. T., & Delimitis, A. (2026). Spectrogram-driven unsupervised autoencoder with isolation forest and one-class SVM for lab-scale wind turbine blade fault detection. Energy and AI, 100681.
3. Capurso, T., Bergamini, L., & Torresi, M. (2022). A new generation of centrifugal pumps for high conversion efficiency. Energy conversion and management, 256, 115341.
4. Chang, K., & Park, S. H. (2024). Random forest-based multi-faults classification modeling and analysis for intelligent centrifugal pump system. Journal of Mechanical Science and Technology, 38(1), 11-20.
5. Chen, L., Wei, L., Wang, Y., Wang, J., & Li, W. (2022). Monitoring and predictive maintenance of centrifugal pumps based on smart sensors. Sensors, 22(6), 2106.
6. Corrales-Bonilla, J., Hidalgo-Osorio, W., Corrales-Otáñez, C., & Viteri-Tapia, F. (2026). A data-driven multivariable framework for operational regime identification, product transition detection, and anomaly detection in industrial pumping systems using SCADA data. Future Technology, 5(3), 150-163.
7. de Souza, D. F., da Guarda, E. L. A., da Silva, W. T. P., Sauer, I. L., & Tatizawa, H. (2022). Perspectives on the Advancement of Industry 4.0 Technologies Applied to Water Pumping Systems: Trends in Building Pumps. Energies, 15(9), 3319.
8. Dias, A. L., da Silva, J. T., Turcato, A. C., & Sestito, G. S. (2021, August). An intelligent fault diagnosis for centrifugal pumps based on electric current information available in industrial communication networks. In 2021 14th IEEE International Conference on Industry Applications (INDUSCON) (pp. 102-109). IEEE.
9. Garousi, M. H., Karimi, M., Casoli, P., Rundo, M., & Fallahzadeh, R. (2024). Vibration analysis of a centrifugal pump with healthy and defective impellers and fault detection using multi-layer perceptron. Eng, 5(4), 2511-2530.
10. Grašs, A. K. (2026). A two-layer approach to predictive maintenance using unsupervised anomaly detection and rule-based fault interpretation: industrial case study of a thermo-oil pump.
11. Johnson, H. A., Simon, K. P., & Slocum, A. H. (2021). Data analytics and pump control in a wastewater treatment plant. Applied Energy, 299, 117289.
12. Khalid, S., Jo, S. H., Shah, S. Y., Jung, J. H., & Kim, H. S. (2024, December). Artificial intelligence-driven prognostics and health management for centrifugal pumps: a comprehensive review. In Actuators (Vol. 13, No. 12, p. 514). MDPI.
13. Marscher, W. D. (2023). Centrifugal pump monitoring, troubleshooting and diagnosis using Vibration technologies. Condition monitoring, troubleshooting and reliability in rotating machinery, 3, 15-76.
14. Martone, A., & Zazzaro, G. (2025). Centrifugal Pump Dataset (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.15301820
15. Martone, A., D’Ambrosio, A., Ferrucci, M., Cembalo, A., Romano, G., & Zazzaro, G. (2025). Sensor-Based Monitoring Data from an Industrial System of Centrifugal Pumps. Data, 10(6), 91.
16. Mohammadi, Z., Heidari, F., Fasamanesh, M., Saghafian, A., Amini, F., & Jafari, S. M. (2023). Centrifugal pumps. In Transporting operations of food materials within food factories (pp. 155-200). Woodhead Publishing.
17. Okirie, A. J. (2024). Evaluation of centrifugal pump availability trends and analysis. Scholars Journal of Engineering and Technology.
18. Shaikh, F., Ahmed, B. S., & Swerin, A. (2025). Unsupervised detection of faults in industrial pumps from multivariate time series. Machine Learning with Applications, 100784.
19. Sunal, C. E., Dyo, V., & Velisavljevic, V. (2022). Review of machine learning based fault detection for centrifugal pump induction motors. IEEE access, 10, 71344-71355.
20. Turk, M. C., Kazemi, Z., Andersen, P. R., Lemming, J., & Larsen, P. G. (2024, August). Implementation of Artificial Intelligence-Based Fault Classification and Anomaly Detection: A Case Study on Hydraulic Centrifugal Pumps. In 2024 IEEE 19th Conference on Industrial Electronics and Applications (ICIEA) (pp. 1-6). IEEE.
21. Ullah, N., Ahmad, Z., & Kim, J. M. (2026). Defect identification in centrifugal pumps based on vulnerable grams and end-to-end deep learning framework. Engineering Applications of Artificial Intelligence, 165, 113412.
22. Zhu, Y., Zhou, L., Lv, S., Shi, W., Ni, H., Li, X., ... & Hou, Z. (2023). Research progress on identification and suppression methods for monitoring the cavitation state of centrifugal pumps. Water, 16(1), 52.
