Research On Control, Monitoring And Diagnosis In Industrial Processes

The control, monitoring and diagnosis for complex industrial process, which are challenging tasks in process control, have attracted increasing attention in control field at present.On the one hand, the complexity of industrial process is continuously rising along with the increasing expansion of production scale. It's not easy to use simple centralized control method or build accurate mathematical model to control these problems. Thus decentralized control strategy or intelligence control method may be one of the best choices of these problems. The decentralized control strategy can simplify the study and analysis of complex problems. And the intelligence control method can settle the problems effectively merely through studying without modeling, which opens up a new approach to solve complex problems.On the other hand, along with the development of demands on control system performance promotes the researches on monitoring and diagnosis in industrial process greatly in recent years. Lots of controllers may have good performance at the beginning of running time in industrial process. However, some time later, the controller performance usually can't meet the requirement of production for the influence such as the environment change or equipment spoilage. If the situation can not be discovered and rectified in time, it will cause great damage to the safety and stability of the industrial production. The purpose of monitoring and diagnosis in industrial process is to recognize and diagnose the controller performance problems in time, and guide the operators to take corresponding schemes to improve the control system performance. Therefore, the discussion and research on monitoring and diagnosis in industrial process have great importance both in theory and application.The main research works and contributions of the thesis are listed as following:1. The history and current research progress of decentralized predictive control, rough control, and control performance assessment and diagnosis are synthetically reviewed.2. Facing to the interior feature of complex large scale system with obvious hybrid dynamics, a multi-timescale decentralized predictive control algorithm based on Nash optimization is presented. The method can adopt different sampling frequencies and control strategies according to the intrinsic characteristics of each sub-system. Therefore, it can completely reflect the control demands of each sub-system. In the meantime, through introducing multi-timescale information prediction and communication, the method can compensate the subsystem for information deficiency caused by different timescale, and improve the control effect.3. The method of simulating Bang-Bang control based on rough set theory is proposed by discussing the similarity between Bang-Bang control and operators' control behavior. In order to extract more sufficient control strategy, extra testing signal is injected to excite the system, which improves the efficiency and maturity of the rough rule extraction. In addition, since the limitation that the rough controller simulating Bang-Bang control causes steady error, an improved method is proposed to combine the rough controller with PID control. The method integrates with the advantages of both Bang-Bang control and PID control, which not only guarantees fast control effect, but also removes steady error.4. The traditional performance assessment methods are difficult to deal with the assessment problem under non-white noise. A method with forgetting factor is presented to solve this problem. It can improve the accuracy of identification and performance assessment under time-variant noise. Simulation results illustrate that this method holds better stability and effectiveness.5. Plant-wide oscillation propagates in the industrial process. A method with closed-loop testing is presented to diagnose the root cause of oscillation. This method can effectively distinguish the oscillatory loops caused by exterior disturbance from the oscillatory loops caused by interior valve friction. The feasibility and efficiency of the proposed method are demonstrated by the simulations.6. Applying the research works based on chapter 4 to the application of PTA production equipment in chemical plant of Yangzi Petrochemical Company Ltd, the scientific research program named "Real-time Performance Monitoring System for Control System Region" were realized. The system passed the plant acceptance check, and performs very well, which not only satisfies the production requirement, but also obtains economic benefit of 4, 100, 000 RMB per year.At the end of this dissertation, some suggestions to the further research in these fields are also given.