基于过程数据分析的彩色滤光片产线故障机台定位

A Location of failure machine in color filter production line using process data analysis

为了完成复杂工艺彩色滤光片产线的故障定位问题,提出一种将加工过程数据和机台警报融合分析的故障定位模型。在该模型中,基于缺陷类型和面积权重评分确定多缺陷共存时的滤光片主缺陷标签,通过Xgboost集成方法训练加工过程数据与缺陷的关联模型,以信息熵增益确定各个过程数据源的贡献度;基于K-means聚类机台警报划分各个故障问题子集,以警报来源反推机台故障概率,进而以D-S证据理论融合两者进行判断,确定综合问题机台概率。通过实例验证了该方法的有效性。

To solve the problem of failure location in complex-process colour filter production line,a failure-locating model combining both processing data and machine alarm for fusion analysis was proposed.In this model,the main defect label of the filter under multi-defect coexistence was determined based on the defect type and the area weight score;the correlation model between processing data and defect were trained with Xgboost integration method,and the contribution of each process data source was further calculated by information entropy gain.Based on the k-means clustering method,machine alarm was divided into several problem subsets,and the probability of machine failure was backward reasoning by alarm source.Then the Dempster-Shafer evidence theory was used to combine the above two judgments to determine the comprehensive probability of the failure machine.The effectiveness of the model was demonstrated by an industrial example.