针对长期正常服役的采掘装备典型故障数据少、有标签数据不足,故障诊断模型训练效果不好等问题,提出一种基于深度迁移学习的采煤机摇臂部传动系统故障智能诊断方法。利用该方法将模拟平台故障数据训练后获取的故障诊断模型参数迁移至采煤机智能故障诊断模型中,从而在不同设备之间进行迁移学习,实现基于小样本数据的采煤机摇臂部智能故障诊断。通过构建预训练的卷积神经网络,将转换为二维时频分布的图像数据集作为预训练模型的输入,并将预训练模型的网络参数迁移至采煤机摇臂传动系统故障诊断模型中,通过保证低层网络不变保留模型的泛化能力,将含有标签的数据集作为采煤机摇臂传动系统智能故障诊断模型的训练数据集对模型进行训练,通过微调高层网络参数进行模型优化和权值更新,得到采煤机摇臂传动系统迁移故障诊断模型,提高了模型的特征提取能力减少了误差。为验证方法有效性,以传动系统滚动轴承为研究对象,采用西储大学轴承数据作为训练集，DDS传动系统平台模拟井下采煤机摇臂部传动系统工况得到滚动轴承监测数据,作为测试集进行试验验证。试验结果表明:滚动轴承平均故障识别精度达到99.59%,与传统的智能故障诊断方法相比,提出的智能故障诊断方法收敛速度快且诊断精度高,能够基于实验室的故障诊断知识,实现高精度设备状态识别与分类。

Aiming at the problems of less typical fault data，insufficient labeled data and poor training effect of fault diagnosis model of mining equipment in long-term normal service，an intelligent fault diagnosis method of shearer rocker arm transmission system based on deep transfer learning is proposed. Using this method，the fault diagnosis model parameters obtained after the training of the simulation platform fault data are migrated to the intelligent fault diagnosis model of the shearer，so as to transfer learning between different equipment，and realize the intelligent fault diagnosis of the shearer rocker arm based on small sample data. By constructing the pre-trained convolutional neural network，the image data set converted into two-dimensional time-frequency distribution is used as the input of the pre-trained model，and the network parameters of the pre-trained model are migrated to the fault diagnosis model of the shearer rocker arm transmission system. By ensuring the generalization ability of the low-level network unchanged，the data set containing labels is used as the training data set of the intelligent fault diagnosis model of the shearer rocker arm transmission system to train the model. By fine-tuning the high-level network parameters to optimize the model and update the weights，the migration fault diagnosis model of the shearer rocker arm transmission system is obtained，which improves the feature extraction ability of the model and reduces the error. In order to verify the effectiveness of the method，the rolling bearing of the transmission system is taken as the research object，and the bearing data of the University of Western Reserve are used as the training set. The DDS transmission system platform is used to simulate the working condition of the transmission system of the rocker arm of the underground shearer to obtain the monitoring data of the rolling bearing，which is used as the test set for experimental verification. The experimental results show that the average fault recognition accuracy of rolling bearing is 99.59%. Compared with the traditional intelligent fault diagnosis method，the intelligent fault diagnosis method proposed in this paper has fast convergence speed and high diagnostic accuracy. It can realize high precision equipment state recognition and classification based on laboratory fault diagnosis knowledge.