Bearing Fault Diagnosis Based on Small Sample Learning of Maml–Triplet

Since the emergence of artificial intelligence and deep learning methods, the fault diagnosis of bearings in rotating machinery has gradually been realized, reducing the high costs of bearing faults. However, in the actual work of the equipment, faults rarely occur, resulting in less fault data. Therefore, it is necessary to study small sample fault data. For the case of less fault data, the Maml–Triplet fault classification learning framework based on the combination of maml and the triplet neural network is proposed. In the framework of Maml-Triplet fault classification, firstly, an initial signal feature extractor is obtained using the Maml training method. Secondly, the feature vectors corresponding to signal data are obtained using depth distance measurement learning in the triplet neural network, and the fault type is judged based on the feature vectors of unknown signal. The results show that the accuracy of the Maml–Triplet model is 2% higher than that of the triplet model alone and 5% higher than that of the Maml–CNN meta learning method. When there are fewer data samples, the accuracy gap is more obvious. Therefore, in the case of less data, the Maml–Triplet model has an excellent fault identification ability.