No abstract

Although there are many literatures on process online fault diagnosis, the moving window of fixed length is used in most of cases. In this study, an approach based on hidden Markov model (HMM) for on-line fault diagnosis in industrial processes is proposed. A variable moving window, the length of which is modified with time, is applied to track process dynamic variables. Before fault diagnosis, the process operating condition is monitored using principal component analysis (PCA) until abnormal operation in the process is detected. Case studies from the Tennessee Eastman plant illustrate that the proposed method is effective. Index Terms -fault diagnosis; hidden Markov model; variable moving window; principal component analysis; Tennessee Eastman plant.

No abstract

A wavelet and hidden Markov model (HMM) based approach is introduced to build the statistical model of process data. Wavelet transform provides a compact, information-rich expression of process data through a set of coefficients that carry localized transient information of process operating condition. The non-Gaussian properties of process data are characterized by a mixture Gaussian distribution. And the serial correlations in the data are described by the state transition of hidden Markov model. Case studies from CSTR illustrate that the inherent characteristics of process data can be accurately modeled by wavelet and HMM.