基于动态贝叶斯网络的智能工厂设备健康评估方法研究

Health assessment for equipment in intelligent factory based on the dynamic Bayesian

针对传统的机械工厂设备健康评估方法存在设备重组困难、主要故障提取无特点以及设备间的交互不明显问题,提出了一种基于动态贝叶斯网络的设备健康评估方法。利用智能工厂生产线上实时采集的数据,对工厂数据进行了特征提取,以每小时产量作为评估工厂健康程度的指标;利用随机森林算法提取了影响该工厂生产线产量的主要故障类型,构建了动态贝叶斯网络,并进行了设备健康评估。研究结果表明:动态贝叶斯网络有着出众的学习能力,当输入样本足够多时,设备健康等级的预估由初始的38.4%上升到了64.1%,先验概率会越来越趋近于真实分布,更有利于分析设备间的交互关系及生产线设备的健康状况;该方法可以对设备进行有效的评估,达到有效维护设备的目的。

In order to address the problem that there were difficulties in equipment reorganization,featureless extraction of major faults and insignificant interaction between devices in the health assessment methods of traditional equipments in factories,a dynamic Bayesian network-based equipment health assessment method was proposed.Using the real-time data collected from the production line of the intelligent factory,the factory data was characteristically extracted and the hourly output was used as an indicator to assess the health condition of the factory;and the hourly output was used as an index to evaluate the health of the factory;the random forest algorithm was used to extract the main fault types that affected the output of the factory s production line,and a dynamic Bayesian network was constructed for equipment health assessment.The results demonstrate that dynamic Bayesian networks have a superior ability to learn,and that when the input sample is large enough,the health level of equipment is estimated to rise from 38.4%to 64.1%,the prior probability gets closer to the true distribution,which is more conducive to analyzing the interactions between equipment and the health of the production line equipment.It can provide an effective way to evaluate the equipment and achieve effective maintenance of the equipment.