物联网环境下的车间生产异常发现与分析

Discovery and Analysis of Workshop Production Abnormality in the Internet of Things Environment

针对离散制造过程中生产异常难以准确评估的问题,提出了一种物联网环境下的车间生产异常发现与分析方法。首先,基于在制品实时状态采集模型,定义了在制品异常事件类型;其次,为了衡量在制品异常事件对车间生产的影响程度,以异常事件发生时生产异常影响因素的状态信息为数据基础,将一维原始数据二维化后,采用一种结合批量归一化和dropout方法的卷积神经网络(Convolutional Neural Networks,CNN)来预测生产任务剩余完成时间,并通过生产任务延迟完成时间来量化车间生产异常程度;最后以某航天车间为案例分析,对所提方法的可行性和有效性进行了验证。

To solve the problem that the production abnormalities in discrete manufacturing process are difficult to evaluate accurately,a method of discovery and analysis of workshop production abnormality in the Internet of Things(IoT)Environment is proposed. Firstly,in view of real-time state Acquisition model of Work-In-Progress(WIP),types of abnormal events of WIP are defined. Secondly,in order to measure the impact of abnormal events on workshop production,a convolutional Neural Network(CNN)model combined with batch normalization and dropout is adopted to predict the remaining completion time of the production task when abnormal events occur,the one-dimensional real-time state information of the factors which affecting the production abnormalities is taken as input of the prediction model after turned into two-dimensional data. The abnormal degree of workshop production caused by abnormal events is quantified by the delay completion time of production task.Finally,the feasibility and effectiveness of the method are verified by the case conducted in an aerospace workshop.