论文部分内容阅读
针对传统的主元分析(PCA)的T~2和平方预测误差(SPE)检验所提供的信息并不一致的缺陷,提出一种改进的PCA方法。该方法采用主元相关变量残差(PVR)和一般变量残差(CVR)统计量代替SPE统计量用于过程监测。将此改进的PCA方法应用到双效蒸发过程的仿真监测,与传统的PCA方法相比,新PCA方法能够有效地识别正常工况改变与过程故障引起的T~2图变化,避免了SPE统计量的保守性,能够提供更详细的过程变化信息,提高了对过程变化的分析与诊断能力。
Aiming at the defects of inconsistent information provided by traditional principal component analysis (PCA) T ~ 2 and square prediction error (SPE), an improved PCA method is proposed. This method replaces SPE statistics for process monitoring with principal component-dependent variable residuals (PVRs) and generic variable residuals (CVRs) statistics. The improved PCA method is applied to the simulation monitoring of double-effect evaporation process. Compared with the traditional PCA method, the new PCA method can effectively identify the changes of T 2 diagram caused by the change of normal working conditions and the process failure, and avoid the SPE statistics The conservative nature of the data provides more detailed information on process changes and improves the ability to analyze and diagnose process changes.