Study On Methodology Of Industrial Process State Supervision

Industrial process state supervision techniques focus on analyzing current states, detecting and diagnosing process anomalies, based on the measurements of process data and characteristics. In this context, whenever an abnormality occurs in a process, the system should respond quickly in detecting and diagnosing malfunctions, or take appropriate actions to control the processes. An efficient supervisory control system that takes advantages of providing safety, steady and optimal conditions is very important for industrial processing.Petri nets are powerful in knowledge representation and inference, as well as in modeling and analyzing complex systems. More recently, Petri nets have been used in fault detection and diagnosis in industrial processes. To develop the applicable techniques of these aspects, this thesis deeply investigated industrial process state supervision methodology using Petri nets. The main contributions of the thesis are as follows:1. Some useful types of Petri nets in the supervision techniques are introduced, including Fuuzy Petri nets, Timed Petri nets and Hybrid Petri nets. An object-oriented concepts was introduced and an Hybrid Petri nets-based modeling and simulation software package with a GUI was developed using VC++;2. Fuzzy Petri nets based knowledge representation and inference methodology for fault propagation and diagnosis are studied. For illustration, an experimental plant is considered, on which the fault diagnosis system has been set up, and some faults occurrence and diagnosis lab works have been done;3. Timing factors are associated with the transitions and the degree of truth of rules in Fuzzy Petri net, a new type of timed Fuzzy Petri net (tFPN) approach for prognostication and diagnosis of abnormal states is proposed. The detailed procedures of abnormal states monitoring based on tFPN models are presented, which are demonstrated through a case study;4. A Petri net supervisor based methodology for characteristic state evolutionary supervision is proposed, with detailed discusses on partitioning hybrid characteristic state space, constructing Petri net characteristic state evolution model and designing Petri net characteristic state controller;5. To deal with the supervision problems of hybrid system featured batch process, a Petri net characteristic state controller based supervisory control system framework is presented. Detailed design and analysis aspects are illustrated with a case study.6. The start-up supervision technology for industrial distillation plant is investigated. Methodology of partitioning start-up control tasks, describing generic control task properties, and modeling using Petri net are presented. A framework of supervisory control system based on Petri net characteristic state controller is implemented.7. Techniques of constructing Petri net model for knowledge base of expert system are studied. An algorithm of developing Petri net knowledge base using Horn clauses is proposed. Finally the architecture of industrial process state supervision expert system are designed and described.