Online control loop performance monitoring for large scale industrial processes

Control performance monitoring (CPM) is the predominant method to maintain and improve production effectiveness in the process industry. However, an existing challenge is to provide on-line performance monitoring for industrial systems which contain different types of control loops. One reason is that the performance monitoring for large-scale systems requires large data storage capacity. In addition, the time complexity of CPM algorithms should be minimized. Furthermore, a number of CPM solutions are designed to work off-line, which demands long-term data collection and human intervention. In this thesis, an online continuous control performance monitoring application for large-scale plants was developed. This application can be used in daily operations as a decision support system. The application was designed to be accurate, scalable, and robust enough to provide fault detection and diagnosis automatically. These detection results are combined with Key Performance Indicators for ranking the overall performance of the plants. The algorithm follows a 3-layer calculation routine which covers 38 performance indices. The algorithm starts by computing the advanced performance indices from basic statistical indices. Afterwards, 10 diagnosis logics are adopted for specific fault diagnosis. Finally, overall performance evaluation is carried out. The usability and accuracy tests verified that the CPM application fulfils the design specifications. It is able to run for a prolonged time period without manual operation, and to provide reliable loop assessment results