【摘要】 针对电磁阀故障识别对专家知识依赖过高，现有智能诊断系统多需要人为提取信号特征等问题，以某型号电磁阀作为研究对象，人为设置故障工况，采集各种工况下的多通道运行数据，利用TensorFlow平台搭建了对该电磁阀的端对端故障识别模型。此外，在此基础上又提出了基于Triplet loss函数的改进模型，并进行了验证测试。结果表明，基于Triplet loss的故障识别模型除具有更高的识别准确率之外，对于在不同动作频率下工作的电磁阀信号有更好的泛化能力。更多还原

【Abstract】 Aiming at the current situation that solenoid valve fault identification relies too much on expert knowledge, or relies on extracting features manually, a certain type of solenoid valve is used as the research object, fault conditions are artificially set, and multi-channel operating data under various operating conditions are collected. Then a fault diagnosis model based on machine learning is built using the TensorFlow platform. On this basis, an improved model based on the Triplet loss function is proposed, and a verification test is carried out. The results show that, in addition to higher recognition accuracy, the improved model has better generalization ability for solenoid valve signals that work at different operating frequencies.更多还原