Application Research Of Neural Network In Distillation Column Control System

The complexity and uncertainty of Distillation column in the product process is thedifficulties in control. The traditional control method can not meet the requirements of thecontrol accuracy for production technology. To this end, the main content of this paper isaiming at the problems existing in the control system of the rectification column, selectingthe propylene rectification tower top temperature control system and the coupling betweentop and bottom temperature as the main content of the research, using neural networkcontrolling approach to a thorough simulation experiment study.First of all, the thesis gives the brief analysis of the rectifying column controlrequirements and interference factors, and then mainly introduces the technologicalprocess of propylene rectification tower, summarizes the difficulties and problems incontrol process.Secondly, my thesis discusses the neural network model with time-varyinguncertainty problems application research in the system. For the problem of the propylenerectification tower top temperature is difficult to get the accurate mathematical model,using RBF neural network to identify the object output. For time-varying problems in itsrunning process, the PID controller, RBF neural network adaptive PID controller isdesigned in two forms and respectively make the simulation experiments, founded thatthe neural network PID controller has better control effect and it can effectively solve thetime-varying problem existing in the process control system.Finally, this thesis discusses application study about the neural network in thepresence of multi-variable, strong coupling, and time-varying system. For propylenerectification tower top temperature and bottom temperature between couplings, we use theneural network decoupling compensation decoupling. And for time-varying problems, PIDcontroller and neural network PID controller are used to control in two ways, respectivelymake the simulation experiments, founded that the neural network PID controller hasbetter control effect.