基于神经网络的热镀锌镀层厚度建模与预测方法

Study on the Modeling and Prediction of Coating Weight in Hot Dip Galvanizing Line Based on Neural Network

连续热镀锌是现代镀锌生产的主要方法,但其镀层厚度系统具有非线性、多变量、时变大滞后等特点,难以建模与控制。本文在对镀锌生产线原理、镀层厚度影响因素分析的基础上,构建与训练镀层厚度预测神经网络模型,该模型具有预测精度高,增益准确,覆盖全工况的优点,为下一步镀层厚度控制系统的开发奠定了基础。

Continuous hot-dip galvanizing is the majormethod of modern galvanizing production. However, the coating weight system is nonlinear process with multi-variables and large time-variant delay, leading to the difficulties both in modelling and control. Through the analysis on the mechanism of hot-dip galvanizing line and the influence factors of coating weight, a Neural Network model is built and trained in this work. The derived model has high accuracy bothin predictionsand gainscovering all the operating conditions, which paves a new way to develop the coating weight control system.