Siamese Neural Network-Based Diagnosis of UAVs Faults Under Small Sample Conditions

A method based on siamese neural networks to diagnose UAVs faults under small sample conditions is proposed for the problem of low data accumulation. To expand the number of samples, a sample pair is constructed by randomly and repeatedly extracting two samples from the faulty samples. The features of the two samples in a sample pair are extracted by an extractor consisting of two one-dimensional convolutional neural networks with shared weights in parallel. The similarity of the features of the two samples is measured by employing Euclidean distance to determine whether they belong to the same class. The proposed method is applied to diagnose UAVs faults under small sample conditions, and the recognition ability of the model is compared un-der different sample sizes. The results show that the proposed method can diagnose UAVs faults under small sample conditions, and the diagnosis accuracy does not decrease significantly with the reduction of sample size.