【作者】 田立勇；

【导师】 毛君； 李晓竹；

【作者基本信息】 辽宁工程技术大学 ， 机械设计及理论， 2020， 博士

【摘要】 滚筒载荷识别与预测是实现采煤机煤岩识别、自动截割及截割部传动系统故障诊断的关键问题。本文通过理论分析、仿真模拟及实验测试相结合的方法,设计发明基于多传感器的滚筒载荷感知方法,构建了多传感器数据特征提取与降噪模型,研究了基于多传感器信息融合的滚筒载荷辨识策略,实现了采煤机滚筒载荷实时感知与精确预测,具体如下:(1)针对采煤机滚筒截割载荷无法获取的问题,制定了基于多传感器融合的采煤机滚筒载荷感知系统总体方案,设计发明截齿载荷测试方法、滚筒扭矩测试方法、摇臂连接销轴测试方法、摇臂变形量测试方法,研究基于多传感器的多参量数据同步采集与传输方法,为滚筒载荷的精确感知奠定基础。(2)针对截割部摇臂壳体和多级齿轮传动系统结构复杂,传感器安装位置无法确定的问题,建立了采煤机截割部传递系统刚柔耦合动力学模型,分析了研究摇臂壳体变形与滚筒载荷间的相互影响关系,通过对比摇臂壳体关键位置的变形规律,得到摇臂前后两侧12个应变传感器最佳安装位置;分析研究了截割部多级齿轮传递系统与滚筒载荷间的相互影响关系,确定了距滚筒端距离最近的齿轮轴6是滚筒扭矩感知传感器的最佳安装位置。(3)针对缩小比例采煤机滚筒截割实验测试结果误差大、精度底的问题,根据采煤机实际结构,研制了截齿三向力、惰轮轴载荷、摇臂连接销轴载荷及摇臂应变测试传感器及数据采集、传输平台,并在张家口煤机厂国家能源煤矿采掘机械装备研发(实验)中心进行1:1模拟井下工况的采煤机滚筒截割实验,获取滚筒工作过程中各传感器的实验测试数据,为多传感器融合滚筒载荷辨识与预测提供支撑。(4)针对滚筒实验数据中包含大量噪声干扰信号问题,构建了基于独立成分和小波分析滚筒测试特征数据提取模型与方法,完成了对各传感器的测试数据进行时域和频域分析,时域分析结果表明:各传感器所得到的测量结果均能体现出滚筒截割载荷的变化规律;频域分析结果表明:各传感器数据的1阶波峰频率均为0.467Hz,为滚筒的回转频率,通过各传感器的各阶频率峰值大小可描述滚筒截割载荷变化。(5)针对单一传感器对滚筒载荷识别测试精度低、稳定性差的问题,以截齿载荷直接测试的滚筒载荷为输出样本,以惰轮轴传感器、摇臂连接轴传感器、摇臂变形传感器测试数据为输入样本,建立基于深度神经网络的滚筒载荷辨识与预测模型,并通过实验数据对预测模型进行验证,验证结果表明预测模型对滚筒三向截割载荷的预测精度达到了83%以上,对滚筒扭矩预测精度可达到95%,说明预测模型具有较高的精度。该论文有图104幅,表16个,参考文献156篇。更多还原

【Abstract】 Drum load identification and prediction are the key problems to realize the coal-rock identification,the cutting automation and the fault diagnosis of transmission system of cutting unit.In this paper,by combining the theoretical analysis,the computer simulation and the experimental test,the drum load sensing method based on multi-sensor is designed,the model for multi-sensor data characteristics extraction and noise reduction is established,the drum load identification strategy based on multi-sensor information fusion is researched,the real-time perception and accurate prediction of shearer drum load is realized.The content of the paper are as follows:(1)Aiming at the problem that the cutting load of shearer drum can not be obtained,the system overall scheme for sensing the cutting load of shearer drum based on multi-sensor fusion is designed,the methods for pick load measuring,for drum torque measuring,for connecting pin shaft measuring and for rocker arm deformation measuring are invented,the multi-parameter synchronous acquisition and transmission based on multi-sensor is researched.The content mentioned above provides the basis for accurate perception of drum load.(2)Aiming at the problem that sensor installation positions can not be determined due to the complexity of the transmission system with multi-stage gears and rocker arm shell of cutting unit,the rigid-flexible coupling dynamics model of transmission system of shearer cutting unit is constructed,the interaction between the deformation of rocker arm shell and drum load is analyzed.By comparing the deformation law for key positions of rocker arm shell,the optimal installation positions for 12 strain sensors located in front and rear sides of rocker arm are obtained.The interaction between the transmission system for multi-stage gears of cutting unit and drum load is studied,the optimal installation position of drum torque sensor is determined to be the gear shaft 6 which is nearest to drum end.(3)Aiming at the problem that the large error and low precision in the experiment of shearer drum cutting with lessen ratio,according to the practical structure of shearer,the pick3-dimensional force sensor,the idler gear shaft load sensor,the load sensor of connecting pin shaft of rocker arm,the rocker arm strain sensor are developed,and the platform for data acquisition and transmission is constructed.The 1:1 shearer drum cutting experiment simulating underground work condition is performed in the Research and Development Center for National Energy Mine Machinery in Zhangjiakou Mine Machinery Co.Ltd.The experiment data of sensors in drum work process are obtained,which provides the basis for the identification and prediction of shearer cutting load based on multi-sensor data fusion.(4)Aiming at the amount noise interference included in drum experiment data,the model and method for characteristics extraction of drum measured data are proposed based on independent components and wavelet analysis.The analysis of sensor data in time-domain and frequency-domain is accomplished.The time-domain analysis result shows that the variation law of drum cutting load can be expressed by the sensor data.The frequency-domain analysis result shows that the first-order wave peak frequency of sensor data is 0.467 Hz,which is exactly the rotary frequency of the drum.The drum cutting load variation can be described by the frequency peaks of each sensor.(5)Aiming at the low measuring precision and poor stability of drum load identification with a single sensor,taking the drum load directly measured from pick load as the output sample,taking the measured data including the data of the idler gear shaft sensor,the data of the rocker arm connecting shaft sensor,the rocker arm deformation sensor as the input sample,the model for drum load identification and predication based on the deep neural network is established,and the model is verified by the experiment data.The verification result shows that the prediction accuracy of the model for drum 3-dimensional cutting load is over 83%,for drum torque is 95%,which shows the prediction model is of high accuracy.There are 104 diagrams,16 tables and 156 references.更多还原