基于人机共融的包装机故障诊断系统的设计与实现

随着工业4.0和物联网时代的来临,基于经验和手册的设备维修方式已不能满足复杂设备维修的要求。而传统的设备诊断系统往往只注重从物理传感器采集数据,缺少引入人的经验,难以拥有自学习能力。本文以ZB45烟草包装机为例,提出一种具有自学习能力的人机共融新型故障诊断系统。系统采用贝叶斯网络,实现对传感器数据的自动推理。通过自然语言处理模块与用户交互,学习用户的维修经验,并用来改进诊断效果。提出了基于凸优化的标签选择方法,根据观察到的现象推荐合适的标签,以快速确定最可能的故障,实现快速找到报警号码对应的故障源。生产现场的实测数据表明,本系统可以有效降低万箱故障次数,有效提高故障诊断精度,降低故障排查时间。

Packing Machine Fault Diagnosis System based on Human Machine Integrated Approach

With the advent of Industry 4.0 and the Internet of Things era, equipment maintenance methods based on experience and manuals can no longer meet the requirements of complex equipment maintenance. However, traditional equipment diagnosis systems often only focus on collecting data from physical sensors, lacking the experience of introducing the human factors, and it is difficult to have self-learning capabilities. Taking ZB45 tobacco packaging machine as an example, a new man-machine integrated fault diagnosis system with self-learning ability is proposed. The system uses Bayesian network to realize automatic reasoning of sensor data. Through the natural language processing module to interact with the user, learn the user"s maintenance experience, and use it to improve the diagnosis effect. A label selection method based on convex optimization is proposed, and suitable labels are recommended according to the observed phenomena to quickly determine the most probable faults, and to quickly find the fault source corresponding to the alarm number. The measured data on the producing spot shows that the system can effectively reduce the number of faults for every 10000 boxes, effectively improve the accuracy of fault diagnosis and reduce the troubleshooting time.