基于Bagging-CVA的动态过程及质量相关故障检测

针对过程数据具有时序相关性以及过程故障是否影响产品质量的问题,提出一种基于Bagging思想和典型变量分析(CVA)的故障检测方法(Bagging-CVA)。采用Bagging思想对建模数据随机抽样构成多组新的数据集,消除数据的时序相关性。分别在每组新的数据集基于CVA方法建立过程相关和质量相关的故障检测模型,同时监测故障对于过程和产品质量的影响。同时提出了一种最优模型选取策略,通过故障检测率和误报率来选出最优模型,降低了传统Bagging方法对多组模型的统计量进行融合的复杂度。通过数值案例和田纳西-伊斯曼过程的仿真实验对方法进行验证。实验结果表明,所提出的改进Bagging-CVA方法可以避免过程数据的时序相关性对故障检测模型的影响,从而提高检测率和降低误报率。此外,还可以进一步分析过程故障是否对产品质量产生影响。

Dynamic Process and Quality-Related Fault Detection Based on Bagging-CVA

A fault detection method Bagging-CVA based on Bagging thought and canonical variable analysis is proposed to solve the problems that the process data with time series correlation and the process faults whether affect the product quality.By using the Bagging method,multiple new training sets are obtained,and the time series correlation of process data is removed.The process-related and the quality-related fault detection models were established by the CVA method in every new training sets.Then the optimal model selection strategy was proposed and the optimal model was selected through the fault detection rate and the false alarm rate,which reduced the complexity of the traditional Bagging method to calculate and merge the statistics of multiple groups of models.Finally,experiments on a numerical examples and the Tennessee Eastman process were used to illustrate the effectiveness of the proposed method.The experimental results show that the improved Bagging-CVA method can avoid the influence of the time series correlation of process data on the fault detection model and improve the fault detection rate and reduce the false alarm rate.In addition,the method can further analyze whether process faults have an impact on the product quality.