数据驱动的热轧带钢头部终轧温度预计算优化

摘要：带钢头部终轧温度(finishing delivery temperature，FDT)的预计算是精轧设定中的一项重要内容。它不仅是带钢全长终轧温度控制的基础，而且关系到轧制速度、轧制力以及辊缝等模型参数的预设定，对成品带钢的质量控制起着关键性的作用。在实际生产中，带钢头部终轧温度主要是通过结合了统计经验和自适应修正的半机理模型来计算，但是由于带钢换热过程的复杂性难以用关系式精确表达，导致了带钢头部终轧温度的预计算精度不高。针对此问题，从数据驱动的角度出发，利用BP神经网络和改进粒子群优化算法（improved particle swarm optimization, IPSO），以半机理模型为主，IPSO-P网络模型为辅，建立了一种混合优化模型。通过仿真实验和实际生产对比，结果表明：相比于单纯的神经网络模型或者半机理模型，混合优化模型的预计算精度和收敛速度均有了很大的提高，达到9667%，具有良好的应用前景。

Pre-alculation optimization of finishing delivery temperature of hot strip head based on data driven

Abstract： The pre-alculation of finishing delivery temperature of strip head is an important content in finishing rolling setting. It is not only the basis of temperature control for the whole length of the strip, but also related to the preset parameters of rolling speed, rolling force and roll gap. It plays a key role in the quality control of finished strip steel. In practical production, the finishing delivery temperature of strip head mainly depends on semi mechanism model which combines statistical experience and adaptive correction. However, due to the complexity of the strip heat transfer process, the temperature cannot be calculated accurately by the semi mechanism model, which leads to the low accuracy of the pre calculation of finishing delivery temperature at the head of the strip. In order to solve this problem, a hybrid optimization model from a data driven point of view which based on the semi mechanism models and supplemented by BP neural network and improved particle swarm optimization (IPSO) algorithm was established. Through simulation experiment, the results show that: compared with a pure neural network model or mechanism model, the pre calculation accuracy and convergence speed of the hybrid optimization model have been greatly improved up to 9667%, and it has a good application prospect.