Industrial Polyester Production Process, Intelligent Control System

The on-line rea-time intelligent control system for the industrial process of PET production of Tianjin Chemical Fibre Factory.is further researched in this paper. Wavelet analysis tools are introduced to carry on data processing and identifying process parameters. Base on further research of CPN fuzzy-neural network, an anthropomorphic intelligent controller is designed to implement real-time intelligent control according to the trend of movement of process curves.In this paper, wavelet analysis tools are used to draw different character of signal and noise in multiresolution space, and also according to the different spread characteristic of the maxims of wavelet coefficient, the detection of patterns of viscosity signal of PET process is carried on. The decomposition and reconstruction of viscosity signals, detection of singularity and fault signals and the filtering of white noise are accomplished using wavelet analysis tools, which prepare data for CPN fuzzy-neural network controller. Such method not only avoid calculation of matrix, reduce amount of calculation, and also in the mean time of acquiring improvement of Signal-to-Noise plus maintain good resolution rate for signal details and insensitivity to the patterns of signal detected.This thesis brings forward a kind of intelligent controller based on Wavelet transform and CPN, which can auto draw control rules. Utilizing the self-learning and memory functions of CPN. The rules are synthesized rapidly, and the changes of control rules are achieved by the changes of partial net weights. Some control rules could be got by experts' experienceand then by on-line study, and information from object being controlled couldbe continuously achieved to modify the control rules. This controller is base on the de-noising dynamical viscosity data curve by Wavelet analysis and singularity detection, predicting the trend of viscosity curve, geting and correcting control rules on-line automatically, impersonating the control method of workers, and finally achieving good online control effect.The study of this thesis provides an effective method for real-time intelligent control of PET process. The results of theoretical analysis and simulation by real industry model show that it is feasible and effective.