基于GMM的间歇过程故障检测  

Fault Detection for Batch Processes Based on Gaussian Mixture Model

作　　者：王静[1] 胡益[1] 侍洪波[1] 

机构地区：[1]华东理工大学化工过程先进控制和优化技术教育部重点实验室,上海200237

出　　处：《自动化学报》2015年第5期899-905,共7页Acta Automatica Sinica

基　　金：国家自然科学基金(61374140)资助~~

摘　　要：对间歇过程的多操作阶段进行划分时,往往会被离群点和噪声干扰,影响建模的精确性,针对此问题提出一种新的方法:主元分析–多方向高斯混合模型(Principal component analysis-multiple Gaussian mixture model,PCA-MGMM)建模方法.首先用最短长度法对数据进行等长处理,融合不同展开方法相结合的处理方式消除数据预估问题;利用主元分析方法将数据转换到对故障较为敏感的低维子空间中,得到主元的同时消除了离群点和噪声的干扰;通过改进的高斯混合模型(Gaussian mixture model,GMM)算法对各阶段主元进行聚类,减少了运算量的同时自动得到最佳高斯成分和对应的统计分布参数;最后将局部指标融合为全局概率监控指标,实现了连续的在线监控.通过一个实际的半导体制造过程的仿真研究验证了所提方法的有效性.A new novel of principal component analysis-multiple Gaussian mixture model （PCA-MGMM） method is proposed in this article to handle the problem about outliers and noise interference, which affects the accuracy of modeling when dividing multiple operation phases in batch processes. At first, a shortest length approach is used to solve the problem of unequal data, while a method of hybrid unfolding of a multi-way data matrix is used to eliminate data estimate problem. Secondly, using PCA sequentially to achieve a low-dimensional representation of the original data space, the operation not only gets the principal but also eliminates the outliers and noise interference. After that the modified algorithm of Gaussian mixture model （GMM） is adopted to automatically optimize the number of Gaussian components and estimate their statistical distribution parameters so as to reduce the computational complexity. Finally, the online monitoring is guaranteed to be continuous by using a global probability index integrated by local probability indices of each operation. The effectiveness and flexibility of the proposed method is validated through an empirical study on a real semiconductor process.

关 键 词：间歇过程 多阶段操作 故障检测 高斯混合模型 全局概率指标 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]