Research On Extended Methods Of PCA

Process monitoring technique based on PCA is one of the most active research topics in the field of process automation and control. However, when applying conventional PCA to industrial process monitoring, a lot of problems appear because of its performance limit. Therefore, domestic and foreign scholars have proposed some improved methods, such as adaptive PCA and MSPCA which extend applications of PCA in the industry process control. This dissertation develops the research based on the predecessor's work. The main research contents are as follows:1. When applying conventional PCA to fault detection, it would lead to false-alarm of the system due to the measured data corrupted with noise and outliers. To overcome the limitations of conventional PCA handling the data corrupted with noise and outliers, an approach is developed by combining the wavelets transform, moving median filter and PCA. This method utilizes the advantage of wavelets transform and moving median filter to preprocess the data to eliminate noise and outliers, reduce and remove the false-alarms. At last, this method is applied to fault detection and has a good effect which proves the method is effective and feasible.2. The process statistical model built by conventional PCA is time-invariant, while real industrial processes are slowly time-varying. To overcome the false-alarm caused by the time-varying process condition, an approach is developed which first utilizes wavelets to eliminate noise and then uses adaptive PCA to update the PCA model recursively by combining the ability of wavelets and adaptive PCA. The simulation result of the process monitoring shows this method can not only reduce the false-alarm points, but also improve the effect of the fault detection.3. A new MSPCA method based on moving median filter is presented. Firstly, preprocessing the original data before PCA to eliminate the outliers by median filtering, and then, by modeling the process data at multiple scales, MSPCA methodology which effectively combined the wavelet transform with PCA is applied to eliminate the non-principal components and small wavelet coefficients. This method can not only improve the detecting ability for small but important changes in data, but also resolve the false-alarms problem caused by the outliers, which is difficult of being dealt with by the existing MSPCA, when measured data contain outliers.