Intelligent condition monitoring for the vertical centrifugal pump using multimodal signals and hybrid models

• Published in: Measurement : journal of the International Measurement Confederation Vol. 242; p. 115813
• Main Authors: Fu, Qiang, Liu, Yingqian, Zhang, Rongyong, Ye, Zhipin, Qin, Haoyang, Tao, Fudong, Li, Huairui, Zhu, Rongsheng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2025
• Subjects: Condition monitoring; Distributed fault diagnosis; Distributed unsupervised anomaly detection; Hybrid models; Multimodal signals; Vertical centrifugal pump; Condition monitoring; Multimodal signals; Vertical centrifugal pump; Distributed unsupervised anomaly detection; Hybrid models; Distributed fault diagnosis
• ISSN: 0263-2241
• DOI: 10.1016/j.measurement.2024.115813

•Utilize anomaly detection and fault diagnosis to achieve condition monitoring.•A Model preselection algorithm is utilized to select anomaly detection models.•Fusion method based on ranking faults’ probability has strong interpretability.•Distributed methods and hybrid models are applied to condition monitoring.•Intelligent condition monitoring research on vertical pumps is conducted. Vertical centrifugal pumps (VCPs) play a crucial role in material transmission and energy conversion in high-risk industries such as nuclear power, oil, thermal power, and shipbuilding. Due to their complex structures and harsh operating environments, VCPs cannot tolerate sudden faults, making precise intelligent condition monitoring (ICM) essential for predictive maintenance. Therefore, we propose an ICM method using multimodal signals (vibration and voltage) and hybrid models, realizing monitoring tasks through distributed unsupervised anomaly detection and fault diagnosis models working in serial. A VCP test rig was built for data collection and method validation. The mean testing accuracy of the anomaly detection and fault diagnosis modules is larger than 0.9, and single faults and composite faults have tested the effectiveness of ICM. The test results show that the proposed method is helpful for the precise ICM of VCPs and can be introduced to the ICM of other complex fluid machinery.