Abstract: In order to accurately diagnose the failure of the reducer of the shearer cutting section, a fault diagnosis method based on Deep Auto-Encoder Networks (DAENs) was proposed. The DAENs model used the oil level in the reducer, oil liquid quality, gear wear loss, cutting motor working temperature, cooling water flow, cooling water pressure and oil moisture as visible inputs. The higher level feature representation was obtained by unsupervised greedy learning. This avoided the complexity and inaccuracy of artificial feature extraction and enhanced the intelligence of identification process. At last, the proposed method is tested, and the results show that the proposed method is faster than the BP neural network, avoids local optimization, and has superior diagnostic accuracy and stability. It can be used for accurate diagnosis of the failure of shearer cutting section reducer.