Harmonic Reducer Fault Detection With Acoustic Emission

The condition of the harmonic reducer is significant for the availability of industrial robots. This condition is easily affected by the high-torque periodic operating environment in the long term. To monitor the condition, a fault detection method, which consists of two unique algorithms, is proposed by utilizing the acoustic emission (AE) signal. The first is wavelet regional correlation threshold denoising (WRCTD), and its function is to reduce the noise contained in the raw acoustic signal. The second is the fusion of operational modal analysis (OMA) and variational mode decomposition (VMD), and its function is to achieve fault detection. Specifically, the parameters of VMD are optimized by OMA-based autoregressive moving average (ARMA) and Pearson correlation coefficient. Then, feature extraction and fault detection are implemented accordingly. In addition, a test rig is established to collect the available data and different operating situations are considered. Comparative experiments are carried out to evaluate the effectiveness of the proposed method. Results show that the accuracy of fault detection could achieve 94.5%, which is better than several typical methods. In addition, the computation time of the proposed method is improved by 8.3%.

Automated summary

The condition of the harmonic reducer is important for the availability of industrial robots. A fault detection method utilizing the acoustic emission (AE) signal is proposed, consisting of two unique algorithms. This method effectively reduces noise and achieves fault detection through feature extraction and signal optimization.