A Study Of Wavelet Entropy In Multivariable Process Monitoring

Fault detection is an important component of industrial process control. Through the production process monitoring, we can effectively guarantee the production safety and product quality stability. The distributed control system (DCS) has been widely used in industrial process. A lot of process data is recorded and preserved. Multivariate statistical monitoring is based on Historical data methods. This method is only depends on the process data and its need not to establish a precise mechanism model. So it has wide application value and theoretical significance. In information theory, entropy is a measure of the uncertainty associated with a random variable. It can Strongly character the system characterization. This thesis mainly information entropy, along with principal component analysis (PCA). In addition, using the information entropy to determine the optimal wavelet basis function and scale. The main research contents include the following:(1) An outline of fault diagnosis methods and classification, and the historical data based on multivariate statistical methods are described. Subsequently, the introduction of the wavelet transform and entropy of the basic concepts and current condition. (2) The mathematical tools which are used in this thesis are described. Including the wavelet transform theory and development status, MCUSUM-MSPCA monitoring method and the calculation of information entropy.(3) The information entropy combined with principal component analysis to monitor the process. The information entropy can characterize characteristics the process signal. Compared with conventional PCA, Entropy-PCA can effectively detect variation and make the process monitoring more reliable and prompt.(4) The information entropy method is applied to MCUSUM-MSPCA. Using different wavelet functions at different scales to decompose the historical data. Then calculate the information entropy of decomposition coefficients. By comparing the information entropy to achieve the automatic selection of wavelet. Compared with MCUSUM-MSPCA,Entropy-PCA greatly simplifying the calculation.