基于神经网络的热轧精轧机组各机架出口比例凸度预报

（1. 北京科技大学冶金工程研究院,北京 100083； 5

2. 北京科技大学国家板带生产先进装备工程技术研究中心）

摘要：针对现场在没有出现可见浪形的情况下，利用 Shohet 判别式设定的出口比例值仍然

预报各机架出口比例凸度。利用神经网络来进行预报，与传统模型相比，可以进行高度的非10

表明，在出口厚度为 2-6mm 的带钢中，出口凸度均处于±10μ m 的误差范围内，从而能够

关键词：热轧; 板形; 比例凸度; 神经网络;

(1. Engineering Research Institute of USTB,University of Science and Technology Beijing; 20

2. National Engineering Research Center of Flat Rolling Equipment, University of Science and

Technology Beijing)

Abstract: Considering noticing no visible shape wave, we still could not guarantee export

proportion crown in the flat aera according to the Shohet discriminant, which showed that the

discriminant could not be applied to the scene. As a result, we make use of neural network to 25

forcast export proportion crown of each stand. Compared with the traditional analytical model, the

neural network has its own advantage, such as highly non-linear simulations, simple and

achievable structure because the import variable for training is large amounts of actual data. The

results showed that the export proportion crown were all in the range of error ±10μm when the

。

采用最多的就是基于 Shohet 判别式的精轧机架比例凸度分配策略，即根据末机架所测40