Multi-Grade Resin Quality Estimation For Industrial Polyethylene Process Based On Adaptive Estimation And Control

In the production of large-scale polyethylene plant, melt index and density are two important quality indicators, according to the needs of the production grade, precise control of melt index and density of the improving efficiency of enterprises is significant. However, due to lack of reliable on-line instrumentation and corresponding controller, large-scale industrial production of polyethylene mainly rely on regular sampling of products for quality indicators control in a long period. Design what can be applied to large-scale polyethylene plant in real-time parameter estimator on the quality indicators and controllers to improve system performance will be significant both in theory and practice.Firstly, due to the lack of on-line analyzer, multi-grade resin quality can not be on-line measured and controlled in the industrial polyethylene process. In this paper, a parameter update law was deduced based on the predictive model of resin quality. According to the off-line lab analytical data, an asymptotic tracking state observer design method is proposed to update the estimation of resin quality and model parameter. In addition, grade transition parameter estimation software was developed under Labview environment. The application results with the proposed method to an industrial Unipol licensed linear low-density polyethylene process verified the feasibility and effectiveness. With this method, industrial PE process can be estimate and advanced process control based on the model of resin quality can be achieved.Secondly, the inverse of non-linear part was added with the traditional advanced control based on the traditional advanced process control and nonlinear separation model theory. The design of nonlinear control problem transfer into an object of general linear control problem, which combined model predictive control and inverse design by state observer.Nonlinear separation model based on the advanced controller is designed to broaden the use of model predictive control.The fluctuation of status in the production process will be eliminate through parameter robust projection, which can still get good rusults even in the deviation of the model.Finally, an impulse response identification method which based on Haar wavelet basis is presented, which can solve the shortcoming of traditional identification of impulse response. As the good localized time-frequency domain characteristic, the function-based of Eykhoff method was replaced by Haar wavelet basis. The simulation results prove that this new impulse response identification method not only ensure the validity of the original Eykhoff methods, but also with the advantage of high recognition accuracy and the ability to overcome system noise.