A Data Mining Based Approach for Gear Fault Diagnostics Using Vibration Sensors

Gears are flight critical components in a helicopter. Effective diagnosis of gear faults plays an important role in preventing the fatal failure, increasing the availability of the helicopters, and reducing the maintenance cost. Extensive research has been conducted on gear fault detection while only limited research on gear fault diagnostics has been reported. In this paper, a data mining based approach for gear fault diagnostics using vibration sensors is presented. The approach utilizes empirical mode decomposition to pre-process vibration signals and extracts time domain fault features as condition indicators for gear fault diagnosis. The time domain condition indicators are used to build a k-nearest neighbor algorithm based fault classifier to identify different types of faults. A case study has been used to demonstrate the effectiveness of the developed approach. Seeded gear fault tests are conducted on a notional split-torque gearbox test rig and real vibration signals are collected. The effectiveness of the presented gear fault diagnostic approach is validated using the gear seeded fault testing data.