A Comparative FEA and MCSA Study of Mixed Eccentricity Faults and their Severity Classification in Brushless DC Motors

Abstract

The proliferation of Brushless DC (BLDC) motors in safety-critical and high-reliability applications necessitates advanced condition monitoring and fault prognosis strategies. Air-gap eccentricity stands as a principal mechanical fault precursor, with its mixed static and dynamic form representing the most probable real-world failure mode. This paper establishes a comprehensive, simulation-driven framework for the diagnosis and quantitative severity classification of mixed eccentricity (ME) faults. A parameterized, high-fidelity 2D Finite Element Analysis (FEA) model of a 1 kW, 4-pole, 24-slot BLDC motor is developed and validated against analytical calculations. The model simulates healthy operation and a spectrum of fault conditions encompassing pure static (SE), pure dynamic (DE), and mixed eccentricity across five severity levels (10% to 50% of nominal air-gap). A detailed analysis of electromagnetic performance degradation is conducted, quantifying increases in torque ripple (up to 220%) and Unbalanced Magnetic Pull (UMP), which exhibits a near-quadratic relationship with eccentricity severity. Subsequently, Motor Current Signature Analysis (MCSA) is applied to the simulated stator phase currents. The study rigorously identifies and tracks the evolution of characteristic fault frequency components at f_s±kf_r​ across all fault types. A novel severity indicator, the Normalized Sideband Amplitude Product (NSAP), is proposed and demonstrates a strong linear correlation (R²=0.987) with the overall eccentricity percentage for ME faults. Finally, a machine learning-based classification pipeline is implemented. A feature vector extracted from the current spectrum is used to train a Support Vector Machine (SVM) model, achieving a 97.3% accuracy in classifying fault severity into four distinct categories (Healthy, Low, Medium, High) on a synthetic dataset. This integrated FEA-MCSA-ML methodology provides a robust, non-invasive blueprint for early fault detection and actionable severity assessment, forming a cornerstone for predictive maintenance systems in BLDC motor drives.