Multivariate state estimation-based condition monitoring of slurry circulating pumps for wet flue gas desulfurization of power plants

The slurry circulating pump, a crucial component of wet flue gas desulfurization (WFGD) systems, often encounters faults such as impeller and bearing box damage due to limestone slurry's intense abrasion, corrosion, and cavitation. The various operating conditions, including frequent switching and rapid frequency variations of the slurry circulating pump will make the working environment worse undoubtedly. To ensure the process safety and operation efficiency of the WFGD system, this study proposed a condition monitoring (CM) method based on a novel multivariate state estimation technique (MSET), which can realize the fault early warning of the slurry circulating pump. To improve the accuracy of the conventional MSET, a memory matrix (MM) construction method based on K -means algorithm is proposed, which can provide a MM that varies with the operating condition of real-time observed vectors. In typical cases of predicting the pump's B -phase current under normal conditions, the root mean square error (RMSE) of our method is 0.25 A, much smaller than that of the conventional MSET method (2.08 A). On the other hand, the computation time of our method remains short (about 20 s) compared to other models (e.g., MEST aided with K -nearest neighbour model needs about 19,000 s) adaptive to varying conditions for over 3000 samples. Also, the proposed CM method can effectively provide early fault warning for impeller damage and bearing box damage of the slurry circulating pump, with 8 days and 1 day in advance, respectively. This can win enough time to inspection & maintenance before troubleshooting.