Studies On Process States Monitoring Technology Based On Timed Colored Petri Nets

Modern process industries are characterized by specific energy flow, material flow and information flow (data and control flow), while production equipments present static physical structure in performances which guides process abnormality propagate along with the movements of the various flows. By means of various measured variable, process abnormal states illustrate causes and consequences of abnormal changes. Inspired by these observances, this paper proposes two novel methods mutually correlated for abnormal state monitoring of industrial processes.An abnormal state usually undergoes a course of emergence, development and propagation which is often implicated in abnormal changes of process pressure, temperature, flow and level, and turns up in terms of process alarms caused by increase or decrease of the process variables. In this context, this paper presents a least square based fitting method for five-element changing profiles, which is able to both identify variable trend modes, such as divergence increasing, convergence increasing, steady running, convergence decreasing and divergence desreasing, and classify process alarms in terms of beyond upper or lower threshold. Therefore, the abnormal information invoved in the operating parameters might be extracted to help judge the extent of abnormality.A new kind of colored timed Petri nets is introduced, in which, the places represent observed variables, the positive and negative transitions describe their causal relationship, the color and time sets associated with places represent five-element trends and duration, respectively, and binary color set tokens describe the alarm beyond upper and lower threshold. Additionally, a place compatibility degree based alarm clustering algorithm is suggested to identify the severity of the abnormal variables, followed by the establishments of a forward reasoning prognosis algorithm and a reverse search diagnosis algorithm for abnormality. Based on these observations, an abnormal state monitoring expert system is developed in Visual Prolog programming language eventually.