Feature-level fusion based on wavelet transform and Artificial Neural Network for fault diagnosis of planetary gearbox using acoustic and vibration signals

in recent years, cause of requiring further reliability and accessibility of the machines, condition monitoring using multi information resources and fusion data resulted rather using one resource has become widely used. In this article, an intelligent system is proposed for fault diagnosis and classification of planetary gearboxes based on fusion the acoustic and vibration data at feature level and using artificial neural network (ANN) classifier. First, acoustic and vibration data of the planetary gearbox was simultaneously collected in four conditions: 1. healthy, 2. Worn, 3. Cracked, and 4. Broken. Gained signals were transmitted from time domain to time-frequency domain by wavelet transform. Then, 30 statistic features were extracted from each one to be used as the classifier input. Artificial neural network was applied as the classifier. The first classification of the faults was based on the extracted features from each sensor where classification accuracy for each acoustic and vibration data was respectively about %88.4 and %86.9. The classification accuracy using fused features was gained as %98.6 indicating the efficiency of proposed method for fault diagnosis the planetary gearbox. Also, %10 accuracy increase gained through using data fusion method clearly put it that using the method can significantly enhance the quality and accuracy of fault diagnosis and as a result condition monitoring of the machineries.