Frequency spectrum-based diagnosis of structural gear faults using dual-sensor acoustic and vibration signals

Identifying gear faults such as tooth loss, cracks, and body damage remains difficult under noisy, single-sensor conditions. This work introduces a dual-sensor fault diagnosis technique that fuses structural vibration measurements from a piezoelectric (PZT) sensor with airborne acoustic emissions measured through a condenser microphone. A single-stage spur gear system is experimentally subjected to a variety of artificially introduced faults, such as tooth loss, crack propagation, and body fractures, during steady-state operation at 1500 rpm. Fast Fourier Transform analysis was performed on both sensor domains to identify gear meshing frequencies , harmonics, and sideband patterns. Quantitative spectral metrics such as peak amplitude, RMS, signal-to-noise ratio, and a newly introduced Kinematic Error Index are calculated for assessing fault severity and progression. Additionally, a composite Fault Severity Index was developed by fusing normalised metrics for fault grading. Sideband development and Principal Component Analysis clustering confirmed the system’s capability in distinguishing fault types with minimal overlap. The PZT sensor detects interior structural faults better than the acoustic sensor, which detects surface-level emissions and meshing effects. Their joint utilisation greatly enhances diagnostic resolution and fault localisation. The proposed framework for real-time gear health monitoring in critical industrial systems is scalable and simple.