中文摘要：

传统的基于多元统计过程监控方法都是假设过程处于单一工况下,而随着进料负荷、产品组分等过程参数的改变,生产过程的工况也随之改变,传统方法便不再适用。针对工业过程中的多工况监控问题,提出了一种基于多模型外部分析和Greedy-KP1M的多工况过程监控方法。首先针对传统外部分析方法描述能力不足的问题,用多模型局部建模代替单一模型来获得更好的描述能力,同时获得监控残差,通过对残差进行监控从而去除多工况的影响,进而将核单簇可能性聚类（KP1M）用于对残差的监控上。该方法拥有和支持向量数据描述（SVDD）相当的监控效果,但计算复杂度却远远小于SVDD。同时,采用Greedy方法提取特征样本,进一步降低了算法计算复杂度。最后将上述方法应用在TE模型和乙烯裂解炉的监控上,结果证明了该方法的有效性。

英文摘要：

Multivariable statistical process control（MSPC）is developed in order to extract useful information from process data and utilize them for process monitoring.But when the change in process feed load or product composition happens,the conventional MSPC method does not function well for the process with multiple operation modes.In order to solve these problems,a novel process monitoring method is proposed based on multiple model external analysis and Greedy-KP1M.First,based on the traditional external analysis,multiple models modeling method is introduced to have a better performance.The multiple operation modes of process are eliminated by multiple models external analysis,and the residual error is got for monitoring.Then,to monitor the residual error,the method called Kernel Possibilistic one-Mean clustering（KP1M）is proposed.KP1M has a good ability to monitor nonlinear process.Its performance is similar with support vector data description（SVDD）.But the computation complexity of KP1M is far less than SVDD’s.Moreover,to reduce the computation complexity furthermore,Greedy method is adopted to extract the feature samples for KP1M modeling.In the end,the proposed method is applied to monitor the TE（Tennessee Eastman）process and the ethylene cracking furnace to show its efficiency.