A Bi-Directional Prediction based on Multi-Stage Deep Multilayer Perceptron for the Productive Process of Polyacrylonitrile Precursor

Xun Cai, Siwei Xue, Tianyou Wang, Yeqian Yang, Jianguo Lv, Kin-Tak Lau, Yanbo Gao

Abstract: The precipitating bath is a vital process of producing Polyacrylonitrile precursor in the production of carbon fiber, involving multi-stages with many independent controlling parameters. In this paper, a bi-directional simulation model of precipitating bath process is proposed, which can not only simulate the whole process of precipitating bath but also predict the controllable parameters of every sub process of precipitating bath. The proposed model consists of two parts: forward cascade multi-level parameter prediction based on multi-layer perceptron and hierarchical backward multi-level parameter prediction based on genetic algorithm. The experimental results show that this model can not only simulate the multistage precipitating bath process but also can predict the process parameters efficiently.

External IDs:dblp:conf/cscwd/CaiXWYLLG24