随着煤矿的智能化技术发展，井下设备运行保障成为煤矿安全生产的重中之重。电牵引采煤机作为煤矿井下开采的重要机电装备，由于运行工况复杂、负载大、维护不到位导致故障频发，直接影响了煤矿生产安全和生产效率。为了保障煤矿井下生产安全，提高生产效率，对采煤机进行有效的健康管理势在必行。摇臂作为采煤机的关键执行机构，采煤过程承受截割载荷，悬臂式窄长壳体应变直接影响内部齿轮传动系统的运行状态，其故障诊断与寿命预测研究需求迫切，对保障采煤机高效安全生产意义重大。

论文提出结合数字孪生和深度学习技术的采煤机摇臂部传动系统虚拟监测与剩余寿命预测方法，通过采煤机摇臂部传动系统常见故障机理及振动特性分析，确定采煤机摇臂部传动系统剩余寿命预测技术方案，深入研究摇臂部传动系统的虚拟监测、故障特征提取与剩余寿命预测方法等内容，为采煤机摇臂部传动系统虚拟监测与寿命预测提供一定的理论支撑。

针对采煤机运行状态监测困难、实时动态分析不足等问题，建立采煤机数字孪生体模型实现虚拟信号监测。以采煤机摇臂部传动系统作为监测对象，结合有限元分析与多体动力学仿真技术建立传动系统联合仿真模型，借助虚拟空间与物理空间之间的数据交互，通过余弦相似度度量虚拟和物理两个信息实体之间的相似性并修正虚拟模型，最终获得正确的虚拟振动信号，实现采煤机摇臂部传动系统在虚拟端的信号监测，为实现虚拟运行状态识别提供数据来源。

针对采煤机摇臂部传动系统运行时，振动信号故障特征提取困难的问题，研究基于一维卷积神经网络的故障特征提取方法，使用实测信号对模型进行训练，并通过网格搜索法获取最优的超参数组合，得到优化后的故障特征提取模型。结合实测信号与虚拟信号进行模型验证，实现虚拟信号的运行状态识别，为采煤机摇臂部传动系统的剩余寿命预测奠定基础。

针对采煤机摇臂部传动系统剩余寿命预测过程中对时间依赖性强的问题，研究基于一维卷积神经网络与双向门控循环单元神经网络相结合的采煤机摇臂部传动系统剩余寿命预测方法，将一维卷积神经网络表征学习获得的特征输入双向门控循环单元神经网络中进行长期依赖关系学习，进而预测设备的剩余寿命。本文使用齿轮退化实验数据验证了该方法的有效性，减少对人工经验和专家知识的依赖，提高剩余寿命预测的准确性。

最后，基于上述的数字孪生体模型与剩余寿命预测网络模型，通过虚实数据交互，完成了基于数字孪生技术与深度学习技术的采煤机虚拟监测与寿命预测系统设计，实现系统的实时状态监测与剩余寿命预测等功能，并结合数字孪生体运行状态与剩余寿命预测值，获取实时、准确的剩余寿命预测结果。实验验证了本文理论研究的正确性，为采煤机摇臂部传动系统剩余寿命系统设计提供了理论支撑。

With the development of intelligent technology in coal mine, underground equipment operation security has become the top priority of coal mine safety production. As an important electromechanical equipment in underground coal mining, electric traction shearer has frequent failures due to complex operating conditions, large load and inadequate maintenance, which directly affects the safety and efficiency of coal mine production. In order to ensure the safety of coal mine production and improve production efficiency, it is imperative to carry out effective health management of shearer. As the key actuator of shearer, the rocker arm bears the cutting load in the process of coal mining. The strain of cantilever narrow shell directly affects the operation state of internal gear transmission system. The research on fault diagnosis and life prediction is urgent, which is of great significance to ensure the efficient and safe production of shearer.

This paper proposes a virtual monitoring and residual life prediction method for the rocker arm transmission system of shearer based on digital twin and deep learning technology. Through the analysis of the common fault mechanism and vibration characteristics of the rocker arm transmission system of shearer, the technical scheme of residual life prediction for the rocker arm transmission system of shearer is determined. The virtual monitoring, fault feature extraction and residual life prediction methods of the rocker arm transmission system are deeply studied, which provides some theoretical support for the virtual monitoring and life prediction of the rocker arm transmission system of shearer.

Aiming at the problems of difficult monitoring of shearer operation status and insufficient real-time dynamic analysis, a digital twin model of shearer is established to realize virtual signal monitoring. Taking the transmission system of shearer rocker arm as the monitoring object, the joint simulation model of transmission system is established by combining finite element analysis and multi-body dynamics simulation technology. With the help of data interaction between virtual space and physical space, the similarity between virtual and physical information entities is measured by cosine similarity and the virtual model is corrected. Finally, the correct virtual vibration signal is obtained, and the signal monitoring of the transmission system of shearer rocker arm at the virtual end is realized, which provides data source for realizing virtual operation state recognition.

Aiming at the problem that it is difficult to extract the fault feature of the vibration signal when the transmission system of the rocker arm of the shearer is running, the fault feature extraction method based on one-dimensional convolutional neural network is studied. The measured signal is used to train the model, and the optimal hyperparameter combination is obtained by the grid search method to obtain the optimized fault feature extraction model. Combined with the measured signal and the virtual signal, the model verification is carried out to realize the operation state recognition of the virtual signal, which lays a foundation for the remaining life prediction of the shearer rocker arm transmission system.

Aiming at the problem of strong time dependence in the residual life prediction process of the shearer rocker arm transmission system, the residual life prediction method of the shearer rocker arm transmission system based on the combination of one-dimensional convolutional neural network and bidirectional gated recurrent unit neural network is studied. The features obtained by the characterization learning of one-dimensional convolutional neural network are input into the bidirectional gated recurrent unit neural network for long-term dependency learning, and then the residual life of the equipment is predicted. In this paper, the effectiveness of the method is verified by the experimental data of gear degradation, which reduces the dependence on artificial experience and expert knowledge and improves the accuracy of residual life prediction.

Finally, based on the above-mentioned digital twin model and residual life prediction network model, through the interaction of virtual and real data, the design of shearer virtual monitoring and life prediction system based on digital twin technology and deep learning technology is completed, and the real-time condition monitoring and residual life prediction of the system are realized. Combined with the running state of digital twin and the predicted value of residual life, the real-time and accurate residual life prediction results are obtained. The experiment verifies the correctness of the theoretical research in this paper and provides theoretical support for the design of residual life system of shearer rocker arm transmission system.

[1]王国法, 刘峰, 孟祥军, 等. 煤矿智能化(初级阶段)研究与实践[J]. 煤炭科学技术, 2019, 47(08): 1-36.

[2]王国法, 张德生. 煤炭智能化综采技术创新实践与发展展望[J]. 中国矿业大学学报, 2018, 47(03): 459-467.

[3]王国法, 王虹, 任怀伟, 等. 智慧煤矿2025情景目标和发展路径[J]. 煤炭学报, 2018, 43(02): 295-305.

[4]管众, 陈会涛, 吕松宝, 等. 突变工况下采煤机截割部齿轮传动系统特性研究[J]. 矿山机械, 2020, 48(09): 64-70.

[5]刘惠, 刘振宇, 郏维强, 等. 深度学习在装备剩余使用寿命预测技术中的研究现状与挑战[J]. 计算机集成制造系统, 2021, 27(01): 34-52.

[6]胡昌华, 张浩, 喻勇, 等. 基于深度学习的复杂退化系统剩余寿命预测研究现状与挑战[J]. 电光与控制, 2021, 28(02): 1-6.

[7]谢娜, 闫顺礼. 基于改进深度置信网络的采煤机摇臂传动系统故障诊断研究[J]. 煤矿机械, 2020, 41(04): 153-155.

[8]Tao F, Qi Q. New IT driven service-oriented smart manufacturing: framework and characteristics[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2017, 49(1): 81-91.

[9]陶飞, 刘蔚然, 刘检华, 等. 数字孪生及其应用探索[J]. 计算机集成制造系统, 2018, 24(01): 1-18.

[10]高士根, 周敏, 郑伟, 等. 基于数字孪生的高端装备智能运维研究现状与展望[J]. 计算机集成制造系统, 2022, 28(07): 1953-1965.

[11]汪雪良, 朱全华, 张涛, 等. 船舶结构监测技术研究进展[J]. 船舶力学, 2022, 26(08): 1246-1253.

[12]宋学官, 来孝楠, 何西旺, 等. 重大装备形性一体化数字孪生关键技术[J]. 机械工程学报, 2022, 58(10): 298-325.

[13]谢嘉成, 王学文, 杨兆建. 基于数字孪生的综采工作面生产系统设计与运行模式[J]. 计算机集成制造系统, 2019, 25(06): 1381-1391.

[14]迟焕磊, 袁智, 曹琰, 等. 基于数字孪生的智能化工作面三维监测技术研究[J]. 煤炭科学技术, 2021, 49(10): 153-161.

[15]李聪波, 孙鑫, 侯晓博, 等. 数字孪生驱动的数控铣削刀具磨损在线监测方法[J]. 中国机械工程, 2022, 33(01): 78-87.

[16]张旭辉, 张超, 杨文娟, 等. 数字孪生驱动采掘工作面远程控制技术分析及发展趋势[J]. 智能矿山, 2020, 1(01): 112-124.

[17]张旭辉, 张超, 王妙云, 等. 数字孪生驱动的悬臂式掘进机虚拟操控技术[J]. 计算机集成制造系统, 2021, 27(06): 1617-1628.

[18]胡伟飞, 方健豪, 刘飞香, 等. 基于数字孪生的掘锚一体机实时状态映射[J]. 湖南大学学报(自然科学版), 2022, 49(02): 1-12.

[19]王学文, 刘曙光, 王雪松, 等. AR/VR融合驱动的综采工作面智能监控关键技术研究与试验[J]. 煤炭学报, 2022, 47(02): 969-985.

[20]张超, 张旭辉, 张楷鑫, 等. 数字孪生驱动掘进机远程自动截割控制技术[J]. 工矿自动化, 2020, 46(09): 15-20+32.

[21]黄祖广, 潘辉, 薛瑞娟, 等. 基于数字孪生的数控设备互联互通及可视化[J]. 制造技术与机床, 2021(01): 128-132.

[22]谭飏, 张宇, 刘丽冰, 等. 面向动力学特性监测的主轴系统数字孪生体[J]. 中国机械工程, 2020, 31(18): 2231-2238+2246.

[23]Zhou Y, Fu Z, Zhang J, et al. A Digital Twin-Based Operation Status Monitoring System for Port Cranes[J]. Sensors, 2022, 22(9): 3216.

[24]Lai X, He X, Wang S, et al. Building a Lightweight Digital Twin of a Crane Boom for Structural Safety Monitoring Based on a Multifidelity Surrogate Model[J]. Journal of Mechanical Design, 2022, 144(6): 064502.

[25]许爱华, 闫俊泉, 伍旭灿, 等. 基于VMD和MP算法的旋转机械故障特征提取[J]. 国外电子测量技术, 2017, 36(08): 11-17.

[26]舒文婷, 程军圣, 黄祝庆, 等. 基于GP-SSD的旋转机械复合故障特征提取方法[J]. 电子测量与仪器学报, 2018, 32(05): 17-24.

[27]杨宇, 李鑫, 潘海洋, 等. 基于非线性模式分解的旋转机械复合故障特征提取方法[J]. 中国机械工程, 2018, 29(24): 2947-2953.

[28]贺王鹏, 孙伟, 苏博, 等. 机械故障诊断的稀疏特征提取方法[J]. 西安电子科技大学学报, 2018, 45(02): 154-159.

[29]曹新城, 陈彬强, 姚斌, 等. 机械故障特征提取的拓扑集分形稀疏字典[J]. 振动与冲击, 2020, 39(06): 210-219.

[30]杨静宗, 施春朝, 杨天晴, 等. 基于CEEMD和RobustICA的机械设备故障特征提取方法研究[J]. 制造技术与机床, 2021(10): 123-127.

[31]Yu X, Dong F, Ding E, et al. Rolling bearing fault diagnosis using modified LFDA and EMD with sensitive feature selection[J]. Ieee Access, 2017, 6: 3715-3730.

[32]Song Y, Liu J, Zhang L, et al. Improvement of Fast Kurtogram Combined with PCA for Multiple Weak Fault Features Extraction[J]. Processes, 2020, 8(9): 1059.

[33]Wang X, Si S, Li Y. Variational embedding multiscale diversity entropy for fault diagnosis of large-scale machinery[J]. IEEE Transactions on Industrial Electronics, 2021, 69(3): 3109-3119.

[34]时培明, 梁凯, 赵娜, 等. 基于深度学习特征提取和粒子群支持向量机状态识别的齿轮智能故障诊断[J]. 中国机械工程, 2017, 28(09): 1056-1061+1068.

[35]赵文清, 严海, 周震东, 等.基于残差BP神经网络的变压器故障诊断[J]. 电力自动化设备, 2020, 40(02): 143-148.

[36]张珂, 王竞禹, 石怀涛, 等. 基于CNN的变工况滚动轴承故障诊断研究[J]. 控制工程, 2022, 29(02): 254-262.

[37]周兴康, 余建波. 基于深度一维残差卷积自编码网络的齿轮箱故障诊断[J]. 机械工程学报, 2020, 56(07): 96-108.

[38]曲建岭, 余路, 袁涛, 等. 基于一维卷积神经网络的滚动轴承自适应故障诊断算法[J].仪器仪表学报, 2018, 39(07): 134-143.

[39]安晶, 艾萍, 徐森, 等. 一种基于一维卷积神经网络的旋转机械智能故障诊断方法[J].南京大学学报(自然科学), 2019, 55(01): 133-142.

[40]刘伟, 单雪垠, 李双喜, 等. 基于并行1DCNN的滚动轴承故障诊断研究[J]. 机电工程, 2021, 38(12): 1572-1578.

[41]Wang D, Guo Q, Song Y, et al. Application of multiscale learning neural network based on CNN in bearing fault diagnosis[J]. Journal of Signal Processing Systems, 2019, 91(10): 1205-1217.

[42]包从望, 江伟, 刘永志, 等. 基于卷积神经网络的采煤机截割部减速器故障诊断研究[J]. 机电工程, 2021, 38(10): 1317-1323+1331.

[43]张小丽, 陈雪峰, 李兵, 等. 机械重大装备寿命预测综述[J]. 机械工程学报, 2011, 47(11): 100-116.

[44]郭亮, 李长根, 高宏力, 等. 大数据背景下基于特征学习的机械设备剩余寿命预测[J]. 西南交通大学学报, 2021, 56(04): 730-735+768.

[45]赵广社, 吴思思, 荣海军. 多源统计数据驱动的航空发动机剩余寿命预测方法[J]. 西安交通大学学报, 2017, 51(11): 150-155+172.

[46]文娟, 高宏力. 一种基于UPF的轴承剩余寿命预测方法[J]. 振动与冲击, 2018, 37(24): 208-213+243.

[47]赵劲松, 张星辉, 贺宇, 等. 基于加窗线性卡尔曼滤波模型的设备剩余使用寿命预测方法[J]. 装甲兵工程学院学报, 2014, 28(02): 26-30.

[48]Li Q, Ma B, Liu J. Remaining useful life prediction of rolling eleme-nt bearings based on different degradation stages and particle filter[J]. Transactions of Nanjing University of Aeronautics and Astronautics, 2019, 36(03): 432-441.

[49]吕明珠, 苏晓明, 陈长征, 等. 基于PCA-UPF的风力机轴承剩余寿命预测方法[J]. 太阳能学报, 2021, 42(02): 218-224.

[50]付洋, 曹宏瑞, 郜伟强, 等. 数字孪生驱动的航空发动机涡轮盘剩余寿命预测[J]. 机械工程学报, 2021, 57(22): 106-113.

[51]朱晓燕, 陈志强. 大数据背景下相似性设备剩余寿命预测研究进展[J]. 科技促进发展,2020, 16(12): 1502-1509.

[52]任博, 董兴辉, 郑凯. 基于相似性的风电机组轴承剩余寿命预测方法[J]. 机械设计与研究, 2016, 32(04): 101-104.

[53]侯晓东, 杨江平, 邓斌, 等. 基于多指标相似性的T/R组件剩余寿命融合预测方法[J]. 国防科技大学学报, 2019, 41(04): 190-198.

[54]Xia T, Shu J, Xu Y, et al. Multiscale similarity ensemble framework for remaininguseful life prediction[J]. Measurement, 2022, 188: 110565.

[55]徐洲常, 王林军, 刘洋, 等. 采用改进回归型支持向量机的滚动轴承剩余寿命预测方法[J]. 西安交通大学学报, 2022, 56(03): 197-205.

[56]马海龙. 基于主元特征融合和SVM的轴承剩余寿命预测[J]. 工矿自动化, 2019, 45(08): 74-78.

[57]李天梅, 司小胜, 刘翔, 等. 大数据下数模联动的随机退化设备剩余寿命预测技术[J]. 自动化学报, 2022, 48(09): 2119-2141.

[58]邱晓梅, 隋文涛, 王峰, 等. 基于相关系数和BP神经网络的轴承剩余寿命预测[J]. 组合机床与自动化加工技术, 2019(04): 63-65.

[59]胡城豪, 胡昌华, 司小胜, 等. 基于MSCNN-LSTM的滚动轴承剩余寿命预测方法[J]. 中国测试, 2020, 46(09): 103-110.

[60]徐萌, 王亚锟. 基于双向长短期记忆网络的DA40飞机碳刹车片剩余寿命预测[J]. 计算机应用, 2021, 41(05): 1527-1532.

[61]丁华, 杨亮亮, 杨兆建, 等. 数字孪生与深度学习融合驱动的采煤机健康状态预测[J].中国机械工程, 2020, 31(07): 815-823.

[62]Ding H, Yang L, Yang Z. A predictive maintenance method for shearer key parts based on qualitative and quantitative analysis of monitoring data[J]. IEEE Access, 2019,7: 108684-108702.

[63]李梦蝶, 赵光, 罗灵鲲, 等. 基于改进CNN-LSTM的飞控系统剩余寿命预测[J]. 计算机工程与应用, 2022, 58(16): 274-283.

[64]车畅畅, 王华伟, 倪晓梅, 等. 基于1D-CNN和Bi-LSTM的航空发动机剩余寿命预测[J]. 机械工程学报, 2021, 57(14): 304-312.

[65]车畅畅, 王华伟, 倪晓梅, 等. 基于改进GRU的航空发动机剩余寿命预测[J]. 航空计算技术, 2020, 50(01): 13-16+20.

[66]杨阳, 米玉泉, 李明, 等. 多源驱动的采煤机短程截割传动系统固有特性分析[J]. 煤炭学报, 2018, 43(11): 3232-3239.

[67]方伟中. 基于改进BP神经网络的采煤机截割部传动系统故障监测与诊断[J]. 煤矿机械, 2020, 41(07): 153-156.

[68]张东升, 于海洋, 徐健博, 等. 采煤机截割部传动系统非线性特性研究[J]. 系统仿真学报, 2018, 30(01): 249-256+265.

[69]陈惠玲, 张福建, 郭岱. 齿轮传动系统动力学研究[J]. 煤矿机械, 2015, 36(04): 139-141.

[70]庄存波, 刘检华, 熊辉, 等. 产品数字孪生体的内涵、体系结构及其发展趋势[J]. 计算机集成制造系统, 2017, 23(04): 753-768.

[71]周林, 郑四发, 连小珉. 加速工况下传动系统扭转振动分析[J]. 振动工程学报, 2010,23(06): 601-605.

[72]Liu X, Huang H, Xiang J. A personalized diagnosis method to detect faults in ge-ars using numerical simulation and extreme learning machine[J]. Knowledge-Based Systems, 2020, 195: 105653.

[73]Liu X, Huang H, Xiang J. A personalized diagnosis method to detect faults in a bearing based on acceleration sensors and an FEM simulation driving support vector machine[J]. Sensors, 2020, 20(2): 420.

[74]张建文, 陈焕栩, 周鹏. 基于小波阈值的局部放电去噪新方法[J]. 电工电能新技术, 2017, 36(08): 80-88.

[75]白鹏. 基于小波分解与SVR组合的风电场短期风速预测[J]. 机械工程与自动化, 2021(03): 89-91.

[76]向家伟. 机械故障数值模拟驱动的生成式对抗网络及智能诊断原理[J]. 中国科学: 技术科学, 2021, 51(03): 341-355.

[77]杨洪柏, 聂昂, 张江安, 等. 基于原始振动信号的往复压缩机卷积神经网络故障诊断[J]. 机械设计与制造工程, 2018, 47(09): 67-70.

[78]宫文峰, 陈辉, 张泽辉, 等. 基于改进卷积神经网络的滚动轴承智能故障诊断研究[J]. 振动工程学报, 2020, 33(02): 400-413.

[79]杨景明, 周豪腾, 杨波, 等. 基于Gabor核的卷积神经网络改进算法及应用[J]. 燕山大学学报, 2018, 42(05): 427-433.

[80]刘立, 朱健成, 韩光洁, 等. 基于1D-CNN联合特征提取的轴承健康监测与故障诊断[J]. 软件学报, 2021, 32(08): 2379-2390.

[81]潘格格. 基于深度迁移学习的采煤机摇臂传动系统故障诊断研究[D]. 西安: 西安科技大学, 2020.

[82]郑小东. 设备全寿命周期管理与故障预测研究[D]. 绵阳: 西南科技大学, 2021.

[83]谢远东, 雷文平, 韩捷, 等. 全矢RNN的轴承故障诊断研究[J]. 机械设计与制造, 2021(09): 27-31.

[84]陈自强. 基于LSTM网络的设备健康状况评估与剩余寿命预测方法的研究[D]. 合肥: 中国科学技术大学, 2019.

[85]葛阳, 郭兰中, 牛曙光, 等. 基于t-SNE和LSTM的旋转机械剩余寿命预测[J]. 振动与冲击, 2020, 39(07): 223-231+273.

[86]车昱娇, 陈云霞, 崔宇轩. KPCA和改进LSTM在滚动轴承剩余寿命预测中的应用研究[J]. 电子测量与仪器学报, 2021, 35(02): 109-114.

[87]梁权. 用于变压器故障诊断深度学习算法优化的研究[D]. 大连: 大连理工大学, 2021.

[88]重庆大学齿轮退化数据. Available from: < https://qinyi-team.github.io/2019/12/Gear-life-cycle-data-download>.