Condition Monitoring is essential for the early fault detection and the enhancement of operational efficiency in centrifugal pumps. Motor current signature analysis (MCSA) is a well-established non-intrusive technique for monitoring motors and driven equipment. However, the monitoring results of the MCSA can be affected by both system faults and variations in supply voltage. In this study, a novel model-based voltage and current analysis methodology is proposed for the torque oscillation detection and condition monitoring of centrifugal pumps. The interconnections between the torque oscillation and stator current response of centrifugal pumps are mathematically described, and the hydraulic torque characteristics related to the pump operation conditions are studied through CFD simulations. The CNN-LSTM-attention framework is utilized to describe the relationship between the current signature and the supply voltage variations of a centrifugal pump operating under healthy conditions. Improved detection of torque oscillation related to pump anomalies is achieved using a monitoring indicator, which is generated based on the comparison between the predicted dynamic threshold and the measured spectrum. Off-design operation, cavitation and impeller damage tests were conducted on a single-stage, single-suction centrifugal pump to validate the proposed methodology. The results demonstrate that the proposed methodology effectively detects signatures when the centrifugal pump deviates from its preferred operation range. The proposed methodology offers easy installation and remote monitoring advantages as it only requires non-intrusive voltage and current transducers. Additionally, the proposed methodology separates the disturbances from the supply voltage, thus providing more sensitive and reliable detection results compared to conventional MCSA techniques.

中文翻译：

---

使用基于模型的电压和电流分析进行扭矩振荡检测并改进离心泵的状态监测


状态监测对于早期故障检测和提高离心泵的运行效率至关重要。电机电流特征分析 (MCSA) 是一种成熟的非侵入式技术，用于监控电机和驱动设备。然而，MCSA 的监测结果可能会受到系统故障和电源电压变化的影响。在这项研究中，提出了一种基于模型的电压和电流分析方法，用于离心泵的扭矩振荡检测和状态监测。从数学上描述了离心泵的扭矩振荡和定子电流响应之间的相互关系，并通过CFD模拟研究了与泵运行条件相关的液压扭矩特性。 CNN-LSTM 注意力框架用于描述健康条件下运行的离心泵的电流特征与电源电压变化之间的关系。使用监控指示器改进了与泵异常相关的扭矩振荡的检测，该指示器是根据预测动态阈值与测量频谱之间的比较生成的。在单级单吸离心泵上进行了非设计运行、气蚀和叶轮损坏测试，以验证所提出的方法。结果表明，当离心泵偏离其首选运行范围时，所提出的方法可以有效地检测特征。所提出的方法具有易于安装和远程监控的优点，因为它只需要非侵入式电压和电流传感器。 此外，所提出的方法将干扰与电源电压分开，从而与传统的 MCSA 技术相比提供更灵敏、更可靠的检测结果。