Study On Data-based Multimode Industrial Process Monitoring Method

In the industrial process monitoring, it is required both to detect process faults as early as possible and to reduce false alarms. Principal component analysis (PCA) and other multivariate statistical process monitoring methods have been widely applied to solve these problems. However, industrial processes often have multiple working modes, and data in different modes show some different statistic characteristics. Multi-mode monitoring problems have been widely studied. In addition, multimode industrial process data often show dynamic characteristics and nonlinear characteristics.In order to solve these monitoring problems, researches have been done in this paper. The main contributions are listed as follows:(1) Based on the researches of related technology and methods, a new multi-mode monitoring method is proposed to solve multi-mode industrial process fault detection problem. The basic idea is that from the cross-mode view, the common underlying characteristics information among different modes is extracted. The common information part reflects common varying characteristics. The specific part is separated from the common part in each data space, and different principal component analysis (PCA) models are established respectively. And multivariate exponent weighted moving average (MEWMA) method is applied to track the dynamic characteristics. By analyzing combination of common part and corresponding specific part, this method can identify the different working modes. When working mode changes, the common part model and the corresponding specific part model are selected for monitoring. In addition, this method and other traditional methods is applied to Continuous annealing process monitoring, and the simulation results show that the proposed method can not only reduce false alarms, but also improve the accuracy of fault detection, Which proposed method has effectively process monitoring capability.(2) The multimode industrial process often has strong nonlinear characteristics. The traditional methods such as PCA method do not work well with non-linear systems. Therefore, by using kernel function method, a kernel multi-mode monitoring method is proposed to improve monitoring capabilities. The fused magnesium furnace industrial process monitoring simulation results show that the proposed multi-mode kernel monitoring method has the ability to deal with the dynamic nonlinear problem, and can reduce false alarms. Moreover, the proposed method can detect faults earlier, which proved the effectiveness and feasibility of the proposed method.