本发明公开了一种基于迁移关系网络的机械故障诊断方法，包括：构建迁移关系网络的源域和目标域数据；构建迁移关系网络样本的训练集和测试集；构建能够检测机械故障类型的迁移关系网络；对迁移关系网络进行训练得到机械故障诊断模型以及对最终的模型进行测试和性能评估。本发明首次提出结合元学习中关系网络与迁移学习的具有Siamese结构的迁移关系网络。利用Siamese结构构造了一个双通道关系网络，分别输入源域全部数据和目标域的无标签数据，增训练时充分考虑了目标域的信息，大大的增加了故障诊断的准确率。MK‑MMD融合到网络中，有效缩小了两个不同领域之间的概率分布距离，使实验室数据运用到实际的机械故障诊断成为可能。

The invention discloses a mechanical fault diagnosis method based on migration relation network, which comprises the following steps: constructing source domain and target domain data of migration relation network; Constructing a training set and a testing set of the migration relational network samples; Build a migration relationship network that can detect mechanical fault types; The mechanical fault diagnosis model is obtained by training the migration relation network, and the final model is tested and evaluated. The invention firstly proposes a transfer relation network with Siamese structure, which combines relation network in meta-learning with transfer learning. Using Siamese structure, a dual-channel relational network is constructed, in which all the data of the source domain and the unlabeled data of the target domain are input respectively, and the information of the target domain is fully considered during training, which greatly increases the accuracy of fault diagnosis. Mmd is integrated into the network, which effectively reduces the probability distribution distance between two different fields and makes it possible to apply laboratory data to actual mechanical fault diagnosis.