Development And Industry Application Of Fractionator Soft Sensor Model

Along with the development of scale and complexity of modern continuous industrial processes, there is an irresistible trend towards using advanced control techniques to achieve steady production and optimal operation. Multi-variables predict control(MPC) is one of the representative strategies. In MPC, some key controlled variables represent products quality may be necessary, which usually are difficult to be measured directly on-line. Furthermore, on-line analytical instruments are very expensive and difficult to maintain, while usually involve considerable time delays. On-line soft-sensing technology is an effective method to deal with this problem, which leads it to be an important tool for the application of advanced control technology.In this thesis, reported research progresses of soft sensor technology are overviewed. By choosing a typical equipment in continuous industry -fractionating column- as research object, the soft-sensing theory are mathematical model and corresponding modelling researched. Based on the first principal modeling theory, a soft sensor model for fractionating column of oil refinery is developed. Moreover, an on-line dual scroll emendation technology is presented to enhance the model computation precision. The above research results have been applied successfully in some refineries.The main contributions and contents of this thesis are described as follows:1. On the basis of analyzing the reported progress of soft sensor technology and the main trend of relative software development, the modelling methods, implementing technology and problems in industrial application of soft sensor technology are discussed. With the typical equipment in process industry-fractionating column- as a research object, the work of this thesis focuses on the first principle model of the technical computation concerning fractionating column and its on-line emendation.2. In terms of the soft sensor modelling of fractionating column, by combining the first principle with the diagram regression modeling methods, we developed 14 computational modules that can on-line compute the property of most products of thefractionating column. These on-line computation modules can give real time estimation of the controlled variables that required by advanced process control strategies. On the other hand, they can provide helpful information for the operators to understand the process, optimize operation, improve product distribution and realize the process monitoring .3. The performance of the soft sensor model for fractionating column is evaluated. In terms of stability and computation precision, the developed model is superior compared with on-line analytical instrument and off-line analyse in laborabary.4. The model of fractionating column was applied to some catalytic cracking units. The practical results prove that the developed model can meet the requirements of industrial application.5. For model maintenance, an on-line dual scroll emendation technique is developed. The effectiveness is studied through its application to maintain a 4-CBA soft sensor model in an industrial PTA oxidation process. Its overpowering superiority compared with other methods has been evidenced in the industrial application. In addition, this technique possesses extraordinary generality and can be exploited in the maintenance of almost all soft sensor models.