Multi-Sensor diagnosis of Rotor Asymmetries in Wound Rotor Induction Machines (WRIMs)

This thesis presents a rigorous experimental study on the condition monitoring of rotor asymmetries in Wound Rotor Induction Machines (WRIMs). By simultaneously evaluating stator current, external stray flux, and mechanical frame vibrations, the research establishes an integrated cross-domain diagnostic framework capable of adapting to complex structural constraints and varying operational regimes, encompassing both direct-on-line (DOL) and variable frequency drive (VFD) supplies. A preliminary focus of this work is the systematic optimization of sensor placement to address the machine's inherent structural anisotropy. Wideband spectral analyses reveal how different mounting locations and measurement axes significantly filter or amplify electromechanical fault signatures. These findings demonstrate that arbitrary sensor placement exposes condition monitoring systems to a severe risk of false negatives, validating the absolute necessity of a spatially optimized hardware setup. The analytical core of the methodology contrasts steady-state Fast Fourier Transform (FFT) analysis with dynamic Short-Time Fourier Transform (STFT) tracking. Experimental evaluations confirm a critical diagnostic limitation under no-load steady-state conditions, where conventional fault indicators are severely obscured by macroscopic spectral leakage across all physical domains. To overcome this intrinsic vulnerability, transient tracking is employed during DOL start-ups, successfully isolating the fault through its kinematic trajectory. However, the introduction of VFDs (V/f scalar control) fundamentally disrupts this approach; constant-slip kinematics and inverter-induced PWM spectral pollution heavily mask the conventional low-frequency fault amplitudes, confining the fault evidence to morphological track distortions known as slip oscillations. To definitively bypass the inverter's masking effects, the dynamic analysis is shifted to the high-frequency spectrum. The research demonstrates that the dynamic tracking of Principal Slot Harmonics (PSH) provides a highly resilient condition monitoring solution, allowing for the clear extraction of fault trajectories even during challenging unloaded acceleration ramps. In this high-frequency domain, the stray flux sensor proves its unparalleled superiority, offering pristine diagnostic signals unaffected by mechanical damping. Ultimately, the results validate a multi-sensor data fusion strategy that successfully bridges empirical observations with state-of-the-art automated trajectory tracking concepts, setting the foundational stage for Deep Learning integration and robust, autonomous predictive maintenance in modern smart-manufacturing ecosystems.