Abstract: Principal Component Analysis（PCA） is a multivariate statistical technique, with a range of applications in data processing and dimensionality reduction. Over the past two decades, PCA method has also been widely applied to various kinds of industrial processes for process monitoring and fault diagnosis with some successes. Due to the increasing volumes of data, process monitoring methods which are based on PCA approaches suffer many limitations, such as great calculation loads and poor real-time performance. In this paper, a new method called Sparse Principal Component Analysis（SPCA） is developed in process monitoring, using the lasso （least absolute shrinkage and selection operator） to produce modified principal components with sparse loadings. And the SPCA can be formulated as a regression-type optimization function to achieve the main elements of choice. Furthermore, the fault detection is then performed by a detection index using model parameters, and the sparse principal component analysis is used in the Tennessee Eastman process（TE processes） monitoring for simulations. Compared with the traditional principal component analysis method, this SPCA approach builds a model based on the sparse modeling data. Therefore it can reduce the amount of calculations and improve the real time performance. As the SPCA model is applied to simulate with real data, the results show that it has better effectiveness in TE processes.

Key words: least absolute shrinkage and selection operator（lasso）, Sparse Principal Component Analysis（SPCA）, state monitoring, Tennessee Eastman（TE） processes

摘要： 主元分析（principal component analysis）是一种多元统计技术，在过程监控和故障诊断中具有广泛的应用。针对过程监控中数据量大的特点，提出一种稀疏主元分析（sparse principal component analysis）方法，通过引入lasso约束函数，构建稀疏主元分析的框架，将PCA降维问题转化为回归最优化问题，从而求解得到稀疏化的主元，并提高了主元模型的抗干扰能力。由于稀疏后主元相关的数据量减少，利用数据建立过程监控模型，减少了计算量，并缩短了计算时间，进而提高了监控的实时性。利用田纳西伊斯特曼过程（TE processes）进行实验仿真，并与传统的主元分析方法进行对比研究。结果表明，新提出的稀疏主元分析方法在计算效率和监控实时性上均优于传统的主元分析方法。

关键词: 最小绝对收缩和选择算子（lasso）, 稀疏主元分析, 状态监控, 田纳西伊斯特曼（TE）过程