New methods for interaction analysis of complex processes using weighted graphs

Abstract

The selection of the structure of a controller in large scale industry processes usually requires extensive process knowledge. The aim of this paper is to report new results on recently suggested methods for the analysis of complex processes. These methods aid the designers in comprehending a process by representing structural and functional relationships from actuators and process disturbances to measured or estimated variables. The methods are formulated in a flexible framework based on graph theory, which can also be used for closed-loop analysis. Additionally, the sensitivity of the methods to scaling and time delays are discussed and resolved. It is also proposed how filtering can be used to restrict the analysis to a frequency region of interest. The feasibility of the methods is shown by the use of three case studies. A quadruple tank process is used to exemplify the methods and their use. Then the methods are applied on a real-life process, the stock preparation plant of a pulp and paper mill. The third study case analyzes a previously published example in closed loop. It is shown that the methods can be used to take efficient decisions on decentralized and sparse control structures, as well as assessing the channel interactions in a closed-loop system.