论文部分内容阅读
针对间歇过程三维数据预处理中不同展开方式的多向偏最小二乘(MPLS)方法在线应用时存在的缺陷,提出改进的MPLS方法。该方法结合传统沿变量展开与批次展开的优势,不仅包含了批次间的信息,在一定程度上去除了过程的非线性及动态性,而且解决了在线应用时数据填充的问题;其次,该方法采用随时间更新的协方差代替固定的主元协方差充分考虑了得分向量的动态特性:最后,引进时变贡献图的故障诊断方法,实现了对故障源的实时跟踪。将该方法应用到工业青霉素发酵过程中,并与传统的MPLS方法进行比较。结果表明:该方法具有更好的监控性能,并能够及时检测故障及跟踪故障源。
Aiming at the defects in online application of multi-direction partial least squares (MPLS) with different expansion modes in three-dimensional data preprocessing of batch processes, an improved MPLS method is proposed. This method combines the advantages of traditional variables expansion and batch expansion. It not only includes the information between batches, but also eliminates the nonlinearity and dynamics of the process to a certain extent, and solves the problem of data filling in online applications. Secondly, The method takes the covariance updated with time to replace the fixed principal component covariance and takes full account of the dynamic characteristics of scoring vectors. Finally, the fault diagnosis method of time-varying contribution graph is introduced to realize the real-time tracking of fault sources. The method was applied to industrial penicillin fermentation process and compared with the traditional MPLS method. The results show that this method has better monitoring performance, and can detect faults and track fault sources in time.