This study investigates the traditional time and frequency domain analysis for vibration signals obtained from harmonic drives operating under different conditions to enhance fault diagnosis in industrial robotics. The theoretical foundations of time-domain and frequency-domain analysis are first systematically examined. Then an experimental platform is established to collect vibration signals under various operating conditions, followed by comprehensive signal processing using statistical values and Fast Fourier Transform (FFT). The results indicate that fault signals exhibit lower RMS values but higher kurtosis than normal signals, with amplitude and frequency fluctuations intensifying under increased load and speed. Unilateral spectral analysis reveals abnormal vibrations near fault characteristic frequencies. Theoretical derivation determined the characteristic fault frequency for flexspline tooth wear to be 49.5098Hz. Experimental spectral analysis reveals that at this characteristic frequency, the amplitude of fault signals substantially exceeds that of normal signals, serving as a distinctive identification feature. The study demonstrates that traditional time-frequency analysis effectively extracts fault features in harmonic drives, offering valuable insights for early fault detection in industrial robotic systems.

Industrial Robots, Fault Diagnosis, Harmonic Drives, Time-Frequency Analysis, Fast Fourier Transform