Coupled vibration model-driven intelligent fault diagnosis in canned motor pumps

Abstract

Canned motor pumps have an unobservable internal structure, making it impossible to directly monitor their operational states or measure the transfer function from internal excitations to casing vibrations. This limitation poses significant challenges in accurately linking internal faults to measurable external signals. To address this issue, this study establishes a coupled vibration model of the canned motor pump to describe the transmission process from impeller hydraulic excitation to casing vibrations. Based on this model, four fault dynamic models are developed to simulate casing vibrations under different fault conditions, supporting fault mechanism analysis. Additionally, a model-based intelligent diagnostic framework is proposed, enabling accurate fault diagnosis under imbalanced data conditions. Experimental results show that the proposed method effectively captures fault characteristic frequency variations consistent with actual conditions. The spectrum of the simulated signals reflects clear physical significance, providing a robust basis for understanding fault mechanisms and improving the reliability and precision of fault diagnosis. This work offers a novel solution for linking internal fault dynamics to external measurements for canned motor pumps, providing a practical potential for handling imbalanced data and advancing fault diagnosis in complex enclosed mechanical systems.