Reliable Fault Diagnosis Method using Kernel Extreme Learning Machine for Gear Failures

Gear transmission is recognized as a kind of extremely important machinery in a wide range of applications. However, due to harsh operation conditions, the gears are prone to failures. Hence, it is crucial to diagnose the gear faults. It is manifested that the vibration analysis is very useful for gear fault diagnosis. But, the gear failures are always complex and reliable intelligent techniques are needed to handle this big problem. To address this issue, a new approach is proposed based on the kernel extreme learning machine (KELM). As a new generation of artificial intelligence, KELM is more powerful than neural network and has strong ability in pattern recognition. We have evaluated the fault detection performance of the KELM in a gear fault test-bed. The detection results indicate that the KELM can detect the gear fault reliably and accurately. Moreover, through comparison with other existing methods, the proposed method produced the best detection rate of 88.3%.