基于神经网络的卷烟包装机运行监测与故障分析研究  

Research on Operation Monitoring and Fault Analysis of Cigarette Packing Machine Based on Neural Network

作　　者：谢崇权[1] 赵一夫 王安宽[1] XIE Chongquan;ZHAO Yifu;WANG Ankuan(Enshi Cigarette Factory,Hubei China Tobacco Industry Co.,Ltd.,Enshi 445099,China)

机构地区：[1]湖北中烟工业有限责任公司恩施卷烟厂,湖北恩施445099

出　　处：《机械与电子》2023年第3期23-26,共4页Machinery & Electronics

摘　　要：为提高卷烟生产产量,同时降低卷烟包装机的维修成本,提出了一种基于聚类方法和神经网络方法的卷烟包装机运行模式监测及故障预测技术。通过引入稳定因子判断包装机的运行模式,当稳定因子小于或等于阈值表示设备处于稳定模式,反之则为过渡模式。通过聚类方法对得到的数据进行处理,设定Hotelling’s T~2检验值与平方预测误差λ_(SPE)为监测的统计值,当统计值超过限值则判断为出现故障,对故障数据进行记录,并通过神经网络方法对下一次故障时间及故障区域进行预测。通过泊松分布计算部件库存容量,在确保库存充足的情况下,在下一次故障之前及时更换机器部件,以避免停机给生产带来巨大影响。最后通过示例分析,证明了该方法的有效性。In order to improve cigarette production and reduce the maintenance cost of cigarette packaging machine,a technology of monitoring operation mode and fault prediction of cigarette packaging machine based on clustering method and neural network method is proposed.The operation mode of the packaging machine is judged by introducing a stability factor.When the stability factor is less than or equal to the threshold,it means that the equipment is in stable mode;otherwise,it is in transition mode.The data obtained are processed by clustering method.Hotelling's T~2 test value and squared prediction error λ_(SPE) are set as the statistical values of monitoring.When the statistical value exceeds the limit value,it is judged that there is a fault.The fault data are recorded,and the next fault time and fault area are predicted by neural network method.The component inventory capacity is calculated by Poisson distribution to ensure adequate inventory and timely replacement of machine parts before the next failure,so as to avoid the huge impact of downtime on production.Finally,an example is given to demonstrate the effectiveness of the proposed method.

关 键 词：卷烟包装机 聚类方法 神经网络 模式检测与故障分析 

分 类 号：TS43[农业科学—烟草工业] TP183[自动化与计算机技术—控制理论与控制工程]