LIVING POLYMERISATION REACTORS: MOLECULAR WEIGHT DISTRIBUTION CONTROL USING INVERSE NEURAL NETWORK MODELS

In principle, it is possible to exercise control over the molecular weight distribution (MWD) of the polymers produced from living polymerisation processes in flow reactors through the control of reactant feeds in a predetermined fashion. Some of the factors that influence the extent to which control can be achieved with feed perturbations to a single stage continuous flow stirred tank (CSTR) reactor have been reported previously. Here, attention is given to the problem of establishing inverse process models as a first step towards a fully automatic control strategy for the synthesis of polymers with pre-ordained MWD in a real process. Particular attention is given to the development of a neural network model for predicting the instantaneous reactor feed conditions for a specified product MWD and characterising the MWD for the purpose of dimension reduction using principal component analysis. Data collected from a simulated ideal reactor process are used in the study. The way in which this approach will underpin a real laboratory-scale polymerisation system is briefly outlined.

[1]  X. Z. Wang,et al.  Software Sensor Design Using Bayesian Automatic Classification and Back-Propagation Neural Networks , 1998 .

[2]  Gregorio R. Meira,et al.  Molecular weight distribution control in continuous “living” polymerizations through periodic operation of the monomer feed , 1981 .

[3]  M. Wulkow,et al.  The simulation of molecular weight distributions in polyreaction kinetics by discrete Galerkin methods , 1996 .

[4]  Ya‐Ping Sun,et al.  Fluorescence of (5,6)-fullerene-C70 in room-temperature solutions: quantum yields and well-resolved spectra as a function of excitation wavelength , 1993 .

[5]  Howard B. Demuth,et al.  Neutral network toolbox for use with Matlab , 1995 .

[6]  Martin A. Riedmiller,et al.  A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[7]  M. Szwarc.  Carbanions, Living Polymers and Electron Transfer Processes , 1968 .

[8]  F. J. Schork,et al.  Reactor operation and control , 1994 .

[9]  S. Qin,et al.  Selection of the Number of Principal Components: The Variance of the Reconstruction Error Criterion with a Comparison to Other Methods† , 1999 .

[10]  Walter C. Merrill,et al.  Neuromorphic learning of continuous-valued mappings from noise-corrupted data , 1991, IEEE Trans. Neural Networks.

[11]  Steven Walczak,et al.  Heuristic principles for the design of artificial neural networks , 1999, Inf. Softw. Technol..

[12]  J. E. Jackson.  A User's Guide to Principal Components , 1991 .

[13]  MSc PhD Xue Z. Wang BEng.  Data Mining and Knowledge Discovery for Process Monitoring and Control , 1999, Advances in Industrial Control.

[14]  R. Lyman Ott.,et al.  An introduction to statistical methods and data analysis , 1977 .

[15]  Kiyotoshi Matsuoka,et al.  Noise injection into inputs in back-propagation learning , 1992, IEEE Trans. Syst. Man Cybern..

[16]  G. R. Meira,et al.  Model reference adaptive control of a continuous polymerization reactor under periodic operation , 1989 .

[17]  C. McGreavy,et al.  Application of wavelets and neural networks to diagnostic system development , 1999 .

[18]  J. Edward Jackson,et al.  A User's Guide to Principal Components. , 1991 .

[19]  R. G. Gosden,et al.  Modeling, Simulation and Control of Polymerization Processes: Some Aspects for Tailored Synthesis , 1999 .