Soft measurement method for melt index of polymer

Abstract

The invention discloses a soft measurement method for a melt index of a polymer. The soft measurement method comprises the steps as follows: determining a first-stage adaptive link hyperplane model, a second-stage static model and a third-stage pulse response model; reading and inputting the online measurement value of an operating variable to the first-stage adaptive link hyperplane model so as to obtain the online real-time estimation value of a partial gas pressure; inputting the online real-time estimation value of the partial gas pressure to the second-stage static model so as to obtain the estimation value of an instant melt index; and inputting the estimation value of the instant melt index to the third-stage pulse response model so as to obtain the estimation value of an accumulated melt index. Based on the advantages of mechanism analysis fusion data modeling and mechanism modeling, the models can adapt to fluctuation in larger range, and the model parameters are more easily obtained at the same time, so that the soft measurement method is suitable for online measurement of a quality index in the polymerization reaction process.

Description

technical field [0001] The invention relates to the field of macromolecules, in particular to a soft measuring method for polymer melting fingers. Background technique [0002] With the development of the economy, the market demand for the diversification of polymer products is getting higher and higher. This requires that the polymerization production process must be able to produce a variety of products, and be able to switch between different products at low cost and quickly. [0003] However, because the polymerization product stays in the reactor for a long time (generally 3-5 hours), the quality index of the reactor outlet product cannot be switched instantaneously, so the switching time is long (generally 4-20 hours), and a large amount of , Unqualified transition products, resulting in greater economic losses. In this switching process, the quality index that is often considered is the melt index, but due to technical or economic reasons, it is difficult to measure...

AI Technical Summary

Problems solved by technology

For the mechanism modeling method, theoretically, the mechanism model can capture the essential characteristics of the dynamic response and describe the process changes in a wide range, but the premise is that a large number of complex differential equations and very accurate physical and chemical parameters are required

However, the establishment of these equations and the estimation of parameters are expensive, and it is often difficult to accurately obtain

Therefore, the soft sensor method based solely on the mechanism model is difficult to be widely used.

[0005] For the purely data-driven method, although it does not require such a complicated modeling process, the parameters can be estimated through the data, but the resulting model does not consider the physical and chemical characteristics and structural information of the object itself, so it can only It is well suited to the existing training data, so it is difficult to be widely used in practical applications

For the polymerization reaction process with uncertainty and strong nonlinearity, it is difficult for the pure data-driven method to adapt to the frequent changes of objects in a large range.

In addition, the advantage of the data-driven method is to establish a system model of the steady-state process with relatively many sample values; for the key dynamic switching process, because the real sample value of the melt index is very scarce, it is difficult to accurately grasp the reaction by the pure data method. dynamic characteristics

Images