Research And Application Of Predictive Control Based On Neural Network PID In The Production Process Of Dangerous Chemicals

With the development of dope-making process, the demand for Dope production technics precision increases constantly, the demand for automatic control also increases constantly. In the practical industrial process control of dope-making, the target often accused of nonlinear, time degenerative and uncertainties. It is difficult to establish their precise mathematical model, the control methods of conventional PID and neural network PID have not met the need of industry. They are generally recognized by the problem of engineering control. To overcome the mismatch, with the backward, non-linear characteristics of temperature control of dope production, the paper introduces the predictive controls based on the neural network PID.As the toxic of free TDI in the dope-making process of polyurethane curing agent, its removal technology is the common concern problems in the international. Developed countries generally use two methods to remove free monomer: one method is solvent extraction, the special disadvantage is the long operational processes, big loss and the other is a film evaporation. As the design of the single separation evaporator which is adapted to a film evaporation is complex. It is very difficult to achieve in the technology of industrialization. There are only a handful of countries to capture the technical. Therefore, the technological is highly secret, there are not public industrialization reportsThe content of free TDI of polyurethane curing agent is generally limited below 0.5% in the developed countries. The technology will enable free TDI of polyurethane curing agent fell to 0.2 percent. The results also open a new path for the removal of other coatings with curing agent in isocyanate monomer and the removal of isocyanate monomer of non-paint industry. In the process of production, through using appropriate technology, to identify the risk of dope, and to eliminate its harm before the accident, people are generally concerned about the problem. As for the control system it self does not require the accused structure forms only emphasized forecast functional characteristics, according to the system of importing-exporting signals to adjust weight of neural network, to make use of rolling strategy continuously optimizes export control and makes its trajectory track expectations. In the paper, it expounds the basal theory and structure of predictive control and neural network, and researches neural network modeling, predictive and control based on the neural network.The feasibility and superior performance can be found out through the emulation of the computer and real application, the paper which brings forward the solution based on predictive control of neural network is obvious and reliable. There is an extensive space to popularize this method involved in this paper to the application.