中文摘要：

针对动态系统模型难以获取的问题，以提高统计监控性能为目标，对现有的动态隐变量法进行了深人研究。首先指出动态隐变量可包含更多的动态信息，但仍具有自相关，为此提出了采用修正控制图的方法对动态隐变量空间进行检测，而对于自相关不显著的残差空间，指出可按照传统方法或是非参数方法建立控制图；接着将过程知识和经验受控平均运行长度的检验考虑在内，给出了一种时滞变量和时滞长度的确定方法；最后，提出了一种根据残差累积和以及递归特征消除算法（recursive feature elimination，RFE）进行故障变量辨识的方法。通过对双效蒸发过程的应用监控，表明了上述方法的有效性。

英文摘要：

To overcome the difficulty of complex dynamic system modeling, conventional dynamic latent variable method is deeply explored to improve the statistical monitoring performance. Firstly, we point out that dynamic latent variables contain more dynamic information than conventional latent variables, but they still have some autocorrelation. Thus, we suggest adopting non-parametric methods to modify the control charts. For monitoring the residual space, corresponding non-parameter methods are also recommended. In the second aspect, taking the process knowledge and empirical in-control ARL（average run length, ARL） verification into account, a method for choosing the lagged variables and the time-lagged length is proposed. The third feature in our research is to propose a new strategy for identifying fault variable, which is based on the cumulative sum of each variable residual and a RFE （recursive feature elimination, RFE） algorithm. The properties of the foresaid methods are verified through monitoring a double-effect evaporator process.