【摘要】 针对单个传感器无法提供掘进机油液系统在长期运行过程中数据互补性和多维性的问题，提出了基于卷积神经网络(CNN)的掘进机多传感器油液系统故障诊断方法。首先，油液监测方法是通过润滑油的理化性能指标和油液携带的磨粒信息反应油液的性能和状态，可以直接从油液数据中学习最好的特征，不需要任何形式的转换以及特征提取。其次，结合多个传感器采集到的数据作为CNN的输入来对油液系统进行故障诊断。最后，将提出的方法与其它机器学习方法在故障状态分类准确性方面进行比较，通过实验验证得到所提出方法的诊断精度高于其它方法，从而实现了高效诊断性能。更多还原

【Abstract】 Aiming at the problem that a single sensor cannot provide the data complementarity and multidimensional of the tunneling machine oil system in the long-term operation process, a fault diagnosis method of the tunneling machine multi-sensor oil system based on convolutional neural networks(CNN)is proposed. Firstly, the oil monitoring method reflects the performance and state of the oil through the physical and chemical performance indicators of the lubricating oil and the wear particle information carried by the oil. It can directly learn the best features from the oil data without any form of conversion and feature extraction. Secondly, the data collected by multiple sensors are used as the input of CNN to diagnose the fault of the oil system. Finally, the proposed method is compared with other machine learning methods in the accuracy of fault state classification. The experimental results indicate that the diagnostic accuracy of the proposed method is higher than that of other methods, thus achieving efficient diagnostic performance.更多还原