Synchronous Machine Fault Detection Using Stray Field Patterns

The patent replaces complex hardware-based frequency analysis systems with software-based FFT processing on simple microcontrollers. The search coil measures magnetic field variations, and the microcontroller performs spectral analysis through FFT algorithms, substituting complex analog signal processing hardware with programmable software solutions that achieve the same fault detection accuracy.

The system uses the synchronous machine's own stray magnetic field emissions as the detection signal source. By analyzing the magnetic field naturally generated by the machine during operation, the system eliminates the need for external excitation sources or complex test equipment, achieving accurate fault detection using the machine's inherent electromagnetic characteristics.

The patent pre-calculates and stores characteristic frequency patterns for various fault conditions (inter-turn short circuits, dynamic eccentricity, etc.) during system development. These predetermined spectral signatures are stored in the microcontroller's memory, eliminating the need for time-consuming reference data collection from healthy machines during operation. During fault detection, the system directly compares measured spectra against these pre-stored fault patterns.

The system creates simplified spectral signature models of fault conditions that capture the essential diagnostic features. Instead of storing complete reference spectra from healthy machines, the patent uses compact representations of fault characteristic patterns that can be quickly compared against measured data, reducing both storage requirements and comparison time while maintaining detection reliability.

The patent develops fault detection patterns based on fundamental electromagnetic principles that are universal across different synchronous machine types. The characteristic frequency signatures of faults like inter-turn short circuits and dynamic eccentricity follow consistent physical laws regardless of machine size or topology. The system uses normalized frequency ratios (e.g., multiples of fundamental frequency) that scale across different power ratings, enabling the same detection algorithms to work universally from small to large machines.

The system adapts to different machine configurations by adjusting key parameters such as fundamental frequency, number of poles, and operating speed, while maintaining the core detection logic. The FFT analysis dynamically adapts to the specific machine's electrical characteristics, and the pattern recognition algorithms adjust threshold values based on machine type, preserving measurement precision across diverse applications through parameter scaling rather than fundamental algorithm changes.