PVC moisture content prediction method based on LSTM deep recurrent neural network

Abstract

The method comprises the following steps: acquiring operation data of a PVC drying system, preprocessing the acquired historical data, sequencing and re-sampling to obtain an operation parameter matrix, recombining the drying operation parameter matrix according to a data format requirement of an LSTM model to obtain a sample set, and randomly selecting samples with a sample size of 80% in the sample set as a training data set and the rest 20% as a verification data set; taking the training data set and the verification data set as input to construct an initial PVC moisture content prediction model based on the LSTM deep recurrent neural network; optimizing the hyper-parameters in the PVC moisture content prediction model based on the LSTM deep recurrent neural network, wherein the optimized hyper-parameters serve as control variables, and training the PVC moisture content prediction model based on the LSTM deep recurrent neural network to obtain a PVC moisture dynamic prediction model; and inputting real-time operation data of the PVC drying system to obtain a PVC moisture content predicted value.

Description

technical field [0001] The invention relates to the technical field of chlor-alkali chemical industry, in particular to a method for predicting moisture content of PVC based on LSTM deep cycle neural network. Background technique [0002] Polyvinyl chloride, referred to as PVC in English, is a polymer formed by polymerization of vinyl chloride monomer (VCM) in peroxide, azo compound and other initiators or under the action of light and heat according to the mechanism of free radical polymerization. . [0003] Polyvinyl chloride is a heat-sensitive, porous white powder with an amorphous structure. Due to its pore structure, polyvinyl chloride tends to absorb a certain amount of water. Excessive moisture content in polyvinyl chloride will lead to agglomeration, and it may lead to mildew for a long time, which not only affects the appearance and color of the product, but also brings many difficulties to subsequent processing; too low moisture content will cause the appearance ...

AI Technical Summary

Problems solved by technology

[0005] There are many factors that affect the drying of PVC. Each factor responds differently to the final moisture content. The factors are coupled with each other, and the pure hysteresis of the drying system is large.

In the existing production process, the operator / automatic control system controls the moisture content of PVC in a relatively "controllable" range through manual experience or traditional control methods. The product moisture content fluctuates greatly, the product quality is difficult to guarantee, and the load changes frequently and cannot be maximized. Maximize the efficiency of production equipment

Images