基于ASGSO-RBF算法的采煤机滚动轴承故障诊断研究

发布时间：2018-03-09 03:06

本文选题：采煤机滚动轴承　切入点：非线性系统　出处：《辽宁工程技术大学》2015年硕士论文　论文类型：学位论文

【摘要】：煤炭资源在我国能源体系结构中具有非常重要的地位和作用,采煤机作为煤矿生产过程的关键设备,是集机械、电子、电气、传动、液压等为一体的复杂机械。采煤机设备的安全、稳定运行对于保证煤炭生产的安全、促进企业生产效率具有重要意义。由于采煤机常处于潮湿、粉尘颗粒多、电磁干扰严重等复杂井下运行环境,时常出现轴承破损等采煤机关键部件故障。一旦出现此类故障,将导致整个煤矿生产过程停滞,乃至瘫痪。针对采煤机滚动轴承故障,本文在深入研究与分析采煤机运行环境、工作特点、影响因素等导致采煤机轴承故障的基础上,提出一种将RBF神经网络(RBF, RBF Neural Network)与自适应步长萤火虫算法(ASGSO, self-Adaptive Step Glowworm Swarm Optimization)相耦合的拟合算法实现对采煤机滚动轴承故障非线性系统的有效辨识。RBF神经网络具备了强大的时变数据处理能力及网络稳定性,因此更能直接表征本质非线性系统的动态特性。以小波包和RBF神经网络为基础,提出了由小波包分解提取各个节点特征能量谱与自适应步长萤火虫算法优化的RBF神经网络进行分类辨识的采煤机滚动轴承故障诊断方法。对振动传感器输出的信号进行小波包分解,运用基于代价函数的局域判别基(LDB)算法对小波包分解进行裁剪,获取最优的特征能量谱,经处理后作为特征向量训练ASGSO-RBF神经网络,建立诊断模型。充分利用经改进后的ASGSO算法强大的全局多目标搜索能力对RBF的权值与中心、宽度在求解空间中进行快速精确的在线搜索,并结合辨识理论建立基于ASGSO-RBF耦合算法的采煤机滚动轴承故障辨识系统。利用井下实际采集到的各影响因素监测数据进行辨识实验,结果表明：在较高学习效率的前提下,其辨识精度和泛化能力明显强于单一的RBF神经网络、GSO-RBF耦合模型以及工程常用的BP神经网络且具有较强的鲁棒性。该方法对井下采煤机故障灾害的防治提供了充分的理论指导。
[Abstract]:The coal resource has a very important position and function in the energy system structure of our country. As the key equipment in the coal mine production process, the coal mining machine is a collection of machinery, electronics, electricity and transmission. The safe and stable operation of shearer equipment is of great significance for ensuring the safety of coal production and promoting the production efficiency of enterprises. Serious electromagnetic interference and other complex downhole operating environment, such as bearing breakage and other key parts of coal mining machine faults. Once such faults occur, the whole coal mine production process will be stalled or even paralyzed. In view of the fault of the shearer rolling bearing, On the basis of deeply studying and analyzing the running environment, working characteristics and influencing factors of the shearer bearing, this paper studies and analyses the bearing failure of the shearer. A fitting algorithm coupled with RBF neural network (RBF Neural network) and adaptive step size firefly algorithm (ASGSO, self-Adaptive Step Glowworm Swarm optimization) is proposed to effectively identify the nonlinear system of rolling bearing fault of shearer. Large time-varying data processing capacity and network stability, Therefore, the dynamic characteristics of essential nonlinear systems can be expressed more directly, based on wavelet packets and RBF neural networks. This paper presents a fault diagnosis method for roller bearing of shearer based on RBF neural network optimized by wavelet packet decomposition and adaptive step size firefly algorithm. Wavelet packet decomposition, The local discriminant LDB (LDB) algorithm based on cost function is used to cut wavelet packet decomposition to obtain the optimal characteristic energy spectrum. After processing, the ASGSO-RBF neural network is trained as a feature vector. The diagnosis model is established. The powerful global multi-objective search ability of the improved ASGSO algorithm is fully utilized to search the weights and centers of the RBF, and the width of the RBF is searched quickly and accurately in the solution space. Combined with the identification theory, the fault identification system of shearer rolling bearing based on ASGSO-RBF coupling algorithm is established. The identification experiment is carried out by using the monitoring data of various factors collected from underground. The results show that: under the premise of higher learning efficiency, Its identification accuracy and generalization ability are obviously stronger than the single RBF neural network GSO-RBF coupling model and the BP neural network commonly used in engineering. This method provides sufficient theoretical guidance for the prevention and treatment of underground shearer fault disaster.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD421.6

【相似文献】