Intelligent Diagnosis of Planetary Gearboxes Based on DAE-CNN

Planetary gearboxes (PGs) are the core transmission components of complex rotating machinery. The diverse forms of internal gear faults often complicate diagnosis and maintenance work. To improve the fault diagnosis accuracy of PGs, this paper proposes a method for diagnosing PG faults based on deep autoencoders (DAE) and convolutional neural networks (CNN). The method fully utilizes the deep learning capability of DAE and combines it with the precise recognition ability of CNN for fault features. Through precise learning in the encoding layer, it achieves deep dimensionality reduction of fault features and further learns these features after dimensionality reduction by CNN. Compared to traditional CNN, the method reduces the computation parameters and resource consumption, resulting in higher precision intelligent fault diagnosis. Experimental results demonstrate that DAE-CNN exhibits stronger fault diagnosis ability than traditional networks.