论文部分内容阅读
在中厚板生产过程中,轧后冷却的温度控制是决定产品组织性能的关键工艺技术。换热系数是温度控制模型的核心参数,由于其影响因素多且复杂,故很难有一个固定的模型来计算。提出一种简单有效的换热系数的自学习模型:先通过参数节点化和插值法得出全范围内的各换热系数影响因素的特征值;再基于k-NN原理,寻找待冷目标钢板与各已冷样本钢板之间的相似度;最后,通过IDW加权平均算法,预估出目标冷却钢板所需的换热系数。现场实际的应用情况表明,使用该自学习模型相比以往可提高控制精度约5%。
In the plate production process, the cooling temperature control after rolling is to determine the performance of the key product technology organizations. The heat transfer coefficient is the core parameter of the temperature control model, because of its many and complex factors, it is difficult to have a fixed model to calculate. A simple and effective self-learning model of heat transfer coefficient is proposed. Firstly, the eigenvalues of the influencing factors of all the heat transfer coefficients are obtained through parameterization and interpolation. Then, based on k-NN principle, And the similarity of each cold sample plate. Finally, the IDW weighted average algorithm is used to predict the required heat transfer coefficient of the target cooling plate. The actual application in the field shows that the self-learning model can improve the control accuracy by about 5% compared with the past.