基于多工况混合模型的故障监控方法研究

摘要

针对多工况过程建立了一个多工况高斯混合模型（Gaussian Mixture Model，GMM），并利用EM （ Expectation Maximization）算法对该GMM参数进行估计。通过引入贝叶斯阴阳算法（ Bayesian Ying-Yang，BYY），实现了GMM中混合工况数目的自动估计。然后，通过在所建GMM的每个分量中构建PCA模型，建立一个多工况故障监控混合模型。最后利用TE过程研究证明了所建模型在过程监控中的有效性。%A multimode Gaussian Mixture Model ( GMM ) was established and estimated with Expectation Maximization ( EM) algorithm, including the mixture mode number associated with the EM algorithm through introducing Bayesian Ying-Yang ( BYY) algorithm.By constructing Principal Component Analysis ( PCA) mo-nitoring model in GMM’ s each component, a multimode fault monitoring mixture model was established.The Tennessee Eastman ( TE) benchmark proves effectiveness of the proposed model in process monitoring.