本发明公开了一种知识驱动的工业机器人智能迁移故障诊断方法及系统,本发明仅使用目标域各个故障类别中的一个标签样本数据,利用样本数据的故障频率先验知识,对输入信号进行带通滤波,提取有效信息,利用极小的代价大大提高了后续诊断的运行速度和准确率。现实中难以获得大量有标注数据,而无标注数据的获取则
相对容易,而无标注数据的获取则相对容易,基于领域自适应迁移网络具有较强的领域适配能力,利用现有带标签的源域数据与无标签的目标域数据进行训练,可以取得很好的诊断结果,使得本发明的诊断方法能够适用于多种工业机器人旋转类零部件的早期故障诊断,显著提高了故障诊断的准确率和效率。
The invention discloses a knowledge-driven intelligent migration fault diagnosis method and system for industrial robots. The invention only uses one label sample data in each fault category of the target domain, uses the prior knowledge of the fault frequency of the sample data, and carries out band-pass filtering on the input signal to extract effective information, thus greatly improving the running speed and accuracy of subsequent diagnosis with minimal cost. In reality, it is difficult to obtain a large number of labeled data, while the acquisition of unlabeled data
It is relatively easy, but the unlabeled data is relatively easy to obtain. The domain-based adaptive migration network has strong domain adaptability. By using the existing labeled source domain data and unlabeled target domain data for training, a good diagnosis result can be obtained, which makes the diagnosis method of the invention suitable for early fault diagnosis of rotating parts of various industrial robots, and significantly improves the accuracy and efficiency of fault diagnosis.