Abstract: Based on the motor current signal method, a mechanical-electrical coupled computational model is established using the actual guide bearing-submersible motor structure. The model starts from an unbalanced oil film, utilizing short bearing theory and incorporating boundary conditions to solve for the internal unbalanced stress of the guide bearing under this condition. The phenomenon of shaft bending in guide bearing misalignment faults is explained. The relationship between the degree of bending and unbalanced stress is provided, and it is pointed out that a bent shaft leads to uneven distribution of air gaps in three dimensions. By analyzing the three-dimensional uneven air gap distribution, the fault characteristics of guide bearing misalignment are located in the spectrum of stator current signals, completing qualitative analysis of the fault features. Through quantitative processing of the fault features, further research is conducted on the quantitative analysis of misalignment fault severity. The relationship between the degree of guide bearing misalignment and the amplitude of fault characteristic frequency components is identified to distinguish general dynamic eccentricity faults, providing a theoretical method to differentiate between the two types of faults. Simulations and experiments validate the effectiveness of the proposed method, which holds practical significance for the fault diagnosis of guide bearings in large-scale submersible motors.