Data-Driven Intrinsic Viscosity Prediction Method For Polyester Fiber Polymerization Process

The textile industry is one of the pillar industries in my country.As the most important textile raw material,polyester fiber accounts for about 80% of the total output of textile materials in my country.Polymerization is the primary process of the polyester fiber melt direct spinning process.The quality of the polymerization process has an important impact on the quality of the polyester fiber product and the subsequent process flow.The intrinsic viscosity is one of the key quality indicators in the polyester fiber polymerization process.Excessive viscosity fluctuation will cause many quality problems such as poor spinning formation,fiber dispersion,and filament winding.At present,the intrinsic viscosity of polyester fiber polymerization process is mainly based on post-event detection.This method has a hysteresis and cannot be involved in the parameter control of the production process in advance,resulting in large losses in the production process and low production efficiency.Intrinsic viscosity is a key quality index in the polymerization process of polyester fibers,and its fluctuation is affected by many process parameters in multiple stages of the polymerization process.At the same time,considering the complex characteristics of high-dimensional time series,time-delay,and nonlinearity of the aggregation process parameters,the current prediction methods are difficult to apply.Therefore,in view of the above problems,this paper conducts the following research on the prediction of intrinsic viscosity during polyester fiber polymerization:1)Aiming at the characteristics of time lag,high dimension of process parameters,strong redundancy between parameters and significant data defects in the polyester fiber polymerization process,a filtering process parameter selection method based on time lag correlation analysis is proposed,and this method outputs a subset of the input parameters of the predictive model.Firstly,the original data set is preprocessed according to the data defects of process parameters;then,for the time lag and high dimension between process parameters,the maximum information coefficient is used as the evaluation criterion for the importance of process parameters,and the maximum cross-correlation time-delay calculation is combined method,to analyze the correlation between process parameters and intrinsic viscosity;secondly,for the redundant variables in the process parameter set,a redundancy analysis method based on approximate Markov blanket is proposed to remove the redundant parameters of the data set;finally,through the method experiment and case experiments to verify the effectiveness of this method.2)In view of the characteristics of many parameters affecting the intrinsic viscosity in the polyester fiber polymerization process,the reaction process is complex,and the data has a nonlinear time series relationship,taking the process parameters after using the process parameter selection method as the input object,an intrinsic viscosity prediction method based on temporal convolution and attention-gated neural networks is proposed.This method mines the internal spatial correlation of high-dimensional time series data features through temporal convolution network,and strengthens the feature extraction ability of the model;it mines multi-dimensional time series data long-term dependency relationship through gated neural network,and at the same time,an attention mechanism is introduced to give gated neural network.The different weights of the hidden layer state of the network model reduce the masking of deep features and strengthen the influence of important time series features.Finally,the effectiveness of the method proposed in this paper is verified by comparing the prediction accuracy of different algorithms proposed in other literatures with actual production data.Finally,taking the polyester production workshop of a polyester fiber production enterprise in Zhejiang Province as the background,the actual demand of the intrinsic viscosity prediction system is analyzed,and the intrinsic viscosity prediction system of the polyester workshop based on microservice architecture is designed and developed.The case is verified,which provides an effective means for controlling the intrinsic viscosity of polyester workshop.