A practical framework for implementing multivariate monitoring techniques into distributed control system

A large number of advanced multivariate techniques have been yet proposed to improve the performance of process monitoring. These advanced monitoring methods are usually implemented on separate hardware/software, which interconnect with Distributed Control System (DCS) using extra communication links. In this paper, the advanced monitoring algorithms are directly implemented and embedded into the DCS structure for the first time. Hence, the need to extra computers connected to the DCS will be eliminated, which brings in several advantages from the performance perspective. In the proposed approach, the advanced monitoring technique is first simplified and split into several functions. Next, Dynamic Link Libraries (DLLs) are created and used for fast execution of different functions in the DCS. Finally, using the generated DLLs and defining functions in WinCC, the monitoring algorithm is executed in real-time. In addition to fault detection, a fault classification algorithm is proposed, which effectively identifies different disturbances in the process. For this purpose, Support Vector Machine (SVM) classifier is used. The advanced monitoring technique is implemented in DCS SIMATIC PCS7 from Siemens (c). The proposed condition monitoring method is tested on a real Gas Pressure Regulating Station (GPRS). The feasibility and effectiveness of the proposed method is experimentally confirmed.