Answers are grounded in this paper and cite their sources.
Demonstrates effective blockage fault detection in centrifugal pumps, suggesting improved industrial reliability.
The blockage in centrifugal pumps arises from blocked filters, debris buildup, contamination, or scaling in impeller flow passages, increasing resistance. This reduces discharge head, flow rate, and efficiency. Prolonged blockage can cause wear, vibrations, cavitation, overheating, degraded performance, and catastrophic failure. In industries such as aviation and nuclear power, early detection through condition monitoring is vital to prevent downtime and losses. An experimental setup simulates blockage faults in centrifugal pumps, including low and high-severity blockages on suction and discharge sides, as well as combined blockages. Data from the microphone, pressure transducers, flowmeter, and accelerometer were collected at a constant pump speed, covering 13 classes: one healthy and 12 fault conditions. This study presents a transformer-based approach for multiclass blockage fault identification in centrifugal pumps using the Patch Time-Series-Transformer (PatchTST) model. The proposed model achieved training accuracy of 99.93% with validation accuracy of 99.49%, the model detected low-severity blockages present on the suction side with 100% and the discharge side with 96.67%. A detailed comparison of the proposed model's performance in blockage fault detection with previous studies is presented. Suction pressure sensor data was concluded as a better indicator for blockage fault detection among other sensors. This work highlights the potential of PatchTST model as an effective tool for blockage fault detection in centrifugal pump systems, contributing to improved reliability and early fault detection in industrial practice.
No takes yet. Share an insight, caveat, or question.
Shinde et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: